As 2025 ends, I want to share this new visualization of the site’s data, showing the AIs improving by month:

Visualizing Massive AI Intelligence Gains

This massive improvement is on a test that doesn’t exist in any training dataset, and which AIs were hopeless at just 1.5 years ago. At that time, it was not obvious that AIs would quickly do so well on the test.

A project for 2026 will be to improve the granularity at the top range of the test (>130) to better detect advancement in the 130-150 zone.

Share

AI Vision Gains

Another new development to highlight: Gemini greatly improved AI vision — the ability of the AI to simply look at the IQ question, without it being described in words — in their November 2025 release of Gemini 3 Pro Preview.

Here’s the curve for that way of testing:

Gemini’s new, consistent score of 116 is a lot better than rivals.

One reason to pay attention to AI “vision” is that it can someday facilitate AIs collecting data themselves, thus learning about the world outside of their limited online training data.

That’s important because it’s possible that AIs are running out of good training data online. But if they can see and understand any image, they could add to their training and start to learn a bit more like we humans do. That applies to them looking at both 2D rendered webpages, or a continuous stream of photos they could take of the 3D real world.

Subscribe now

Chinese AIs Gain at Similar Pace

TrackingAI.org recently started tracking Kimi K2 Thinking, which is a Chinese open-source model. Open-source means that the code is publicly available, so anyone could run the AI from their own server — in contrast to, for example, ChatGPT, where you access it by paying OpenAI to get answers from their server.

At an IQ of 116 on the test (when the problems are described in words) it is by far the best open-source model tested to date:

This also suggests that Chinese AI developers are running in pace with American developers, because in Jan 2025, Deepseek scored 100, compared to 118 for the leading US AI at that time. Now, Kimi K2 scores 116, and the leading US AI scores 130 — a similar difference.

Share

Grok’s turn in the sun

For the first year of Grok’s existence, it was one of the worst-performing AIs, and never broke an IQ of 70 in our average.

For much of 2025, it remained a laggard, not breaking 100 up through July.

But then in August, things changed rapidly, and Grok briefly took the #1 spot for a time. As 2025 ends, Grok is in the top 3, closely clustered with Gemini and OpenAI:

It’s great that there is a diversity of smart AIs, so that no one company is controlling this new tech.

Elon Musk also recently tweeted data sourced from TrackingAI.org:

I’ll note again that, because the test isn’t online, it’s not super easy for any company to game the metric. Of course, one could try to teach an AI how to do IQ tests generally, but that should not be particularly easy, given the complex logic in play. For some examples of the kinds of questions used, see this post.

Concluding 2025

I should also mention that AI — or rather, “the architects of AI” — won TIME person of the year in 2025:

Both businesses and individuals are still gradually learning where AI can, and can’t yet, be useful in their work. The moment we’re in is very much comparable to the 1990s when it comes to the internet. The road ahead remains unpredictable, just as it would have been hard to predict TikTok or bitcoin in 1998. We’ll continue to track developments here, and at TrackingAI.org.

Thanks for reading, and Happy New Year!

Did that have an impact on crime?

I haven’t seen anyone else analyze that, so I pulled up the numbers.

Below, the gray lines show cumulative DC homicides in 2024. The blue lines show the same for 2025, up until August 11. The green lines show what happened after National Guard deployment on August 11.

On its face, the graph shows a clear fall in DC homicides starting around when the National Guard was deployed, and the decline has continued through November.

Homicide data is generally the most reliable data, as underreporting is unlikely to be a major problem (in contrast, data for lesser crimes also reflect whether a crime victim thinks police will actually do anything about their stolen bike or graffitied door.)

Using linear regression on the 2025 pre-intervention months (that’s the dotted blue line above) suggests that, if pre-August 11 trends had continued, there would’ve been 112 homicides by September, 126 by October, and 136 by November 4.

Instead, we’ve seen 118 by November 4.

This suggests that approximately 18 lives have been saved due to federal action, and that the impact is ongoing.

Subscribe now

Visually, it’s the gap between the dotted blue line and the last green bar.

I like the cumulative graph above, because it helps visualize the trendline — but here’s what the raw homicide numbers look like:

Could the decline in murders instead be attributed to some nationwide decline that began at the same time? To get a sense, I pulled up data for NYC, as that city also offers nearly-live data:

We see no post-August decline in NYC, which is in line with the hypothesis that the federal DC actions caused a crime reduction.

In a couple of months, tools like the “Real Time Crime Index” will provide preliminary nationwide data, but for now, NYC gives us that quick sanity check.

What about violent and property crime?

As noted, crimes less serious than murder can suffer from reporting issues. But I took a look anyway.

Here’s violent crime:

A linear regression projects 2,160 violent crimes reported by Nov 4, based on pre-intervention trends. Instead, there were 2,140. So basically no difference.

For property crime, we get a similar result: There would’ve been 18,877 property crimes in DC, had pre-August 11 trends continued. Instead, there were 18,795. So, again, slightly lower but basically the same trend.

It is possible that actual crime fell, but was compensated for by an increase in willingness to report crimes. However, we don’t have the data to determine that.

Share

Implications

The visible decline in homicides that aligns with federal action in DC suggests that something worked, saved about 18 lives, and is on track to save more.

Regarding the 18 lives saved so far, that works out to about one life every 5 days over the 85-day period since the federal actions in DC began. Media reports claim that the cost of National Guard deployment in DC is about $1.8 million a day, and some federal government agencies have, in recent years, put the value of a life at about $13.7 million.

Those numbers suggest that the federal actions pass a cost-benefit test, when looking purely at monetary cost and lives saved.

It also suggests that law-and-order is undersupplied generally, when even an emergency action passes such a cost-benefit test. If local and state governments were to take street crime seriously by hiring more police, prosecutors, and judges, that could be even more cost-effective in the long run. It would also have compounding positive effects, as broader culture started to reflect that fact that crime doesn’t pay.

The data don’t make clear whether the shift was due to the National Guard deployment (which allows police resources to be freed up, among other things), the increase in prosecutions, or another policy aspect. Perhaps all those things contributed to an environment where criminals wanted to lay low and not murder people.

Comments are turned off, but feel free to message with your thoughts and ideas:

The New Yorker wonders, “What if A.I. Doesn’t Get Much Better Than This?” while the WSJ reports that “Large language models’ pace of improvement has moderated...”

Through the data generated by my site, TrackingAI.org, I can quantify AI progress on intelligence. Last year, the AI intelligence curve looked like this:

Today, that graph — using the same test, which is “offline”, meaning it’s never been in any AI training data — looks like this:

So there’s been massive progress in AI intelligence.

But what does progress look like when it’s broken down by month? Can we see a slowdown there?

Here’s a new graph, showing the score of the top model over time, by month:

In about 1.5 years, the top AI IQ went from the mid-80s to about 130, which is roughly the difference between the typical high school dropout and someone who’s getting a math degree in a university.1

From May 2024 to October 2025, the leading AI improved by an average of 2.5 IQ points per month, consisting of relatively steady, linear, incremental breakthroughs.

We also see major progress as recently as August 2025.

However, OpenAI’s much anticipated ChatGPT “5” advance instead ended up being mostly a renaming of their older models (though it did contain one major breakthrough; more on that below.) I think that fooled some pundits into thinking that AI as a whole isn’t progressing.

Meanwhile, Elon Musk’s Grok took the lead by raising the top model from about 120 to 130. Anthropic’s Claude 4 Opus also, in the last week, pulled into a tie with Grok.

Subscribe now

Other measures of IQ show similar progress

My site tracks four different measures of IQ.

In addition to the “offline” test described above, there’s the Norway Mensa test. Here’s the AI progress on that:

The smartest AIs have already essentially beaten this test. Twice, ChatGPT Pro has scored 34/35 questions, giving it a score of 148.

I can’t show the questions for the “offline” test, since that’d give them away to future AIs, but I can show some of the Norway Mensa problems to give you a sense of them. Here’s the easiest question, which every AI gets right:

To answer that, one just needs to see that the outer shape retains its pattern across each row, but that the inner shape changes in a predictable pattern. The answer is E.

In contrast, the hardest question is the one below, and while I’m not sure what the pattern is, Grok-4 Expert Mode and ChatGPT 5 Pro nailed it:

The Norway Mensa test is public, and if one looks hard, it is possible to find answers for the questions online. That said, the AIs aren’t told anything about the test, and are merely presented descriptions of these puzzles. So it’s hard for them to simply Google the questions. Still, the questions might be in the AI training data in some form.

That’s probably why the top AIs ace the Norway Mensa test, and are near the 150 maximum, while they score a lower ~130 on the “offline” quiz.

AIs are also improving on critical “vision” IQ tests

Another really interesting test is “vision” IQ tests, which are directly comparable to how we humans take the tests.

In the above charts and tests, the puzzles were described to the AIs in words.

That’s because, back when I first tried testing AI IQs in February 2024, I initially gave them the images themselves, and the leading AI at the time was hopeless. Its results were barely distinguishable from chance, which would give it an IQ score in the 60s. Describing the problems gave the AIs a chance to show off their thinking abilities, even though they were essentially blind.

But about a year ago, the AIs started getting good enough at seeing pictures that they could answer some questions using just the images themselves, without text descriptions. I label those “vision” models.

Here’s the month-by-month performance when the test is administered that way, for the offline test:

ChatGPT Pro’s performance on the offline vision test — averaging 105 — is right in line with the average Maximum Truth reader, who scored 104 on the same questions!

And here’s the Mensa Norway test, given to the AI visually:

ChatGPT 5 made exactly one breakthrough that I was able to see, and that was in terms of vision.

In general, top AIs are scoring respectably by this measure, getting between 105 and 112, which is around the IQ of the typical American college student.2

On the offline vision test specifically, there is indeed a plateau since April 2025. But if we smooth it out and treat it linearly, it’s been gaining an average of 2 IQ points per month since May 2024.

Projecting that out, we should see leading AIs begin to ace the test in about 2 years. I’ll be sure to check back on this prediction on Halloween 2027.

The evidence from the non-vision-test described earlier shows that AI is fundamentally capable of this kind of reasoning; it’s just a matter of training the AIs visually.

Speaking of that, it’s interesting to observe that while Grok leads the field in pure reasoning, it is nearly blind:

Perhaps Grok’s coders just haven’t been training it very much on visual inputs yet.

Subscribe now

IQ is measuring something real

In my own use of AIs, I see a strong correlation between the AIs as measured by my tests, and their performance on real-world questions.

As one funny example, I saw Scott Alexander highlight this yesterday:

… if you ask questions like “answer with a single word: were any mammoths still alive in December”, chatbots will often answer “yes”. It seems like they lack the natural human assumption that you meant last December, and are answering that there was some December during which a mammoth was alive.

Slightly surprised, I flipped open Claude Sonnet (IQ 111 on the offline non-vision test) and was able to replicate it:

But then I tried ChatGPT Pro (IQ 117) and Claude Opus (IQ 121) and they both correctly answered “no.”

Share

Implications of AI getting smarter at a linear rate

These data suggest that AIs continue to get smarter, and at a manageable and linear-ish rate. The field of AI may or may not be over-invested in, but the data hint that we should expect further advances.

I don’t think this tells us much yet about whether AI will someday conquer the world, kill everyone, make us all live forever, or whatnot. But based on my data, I expect AIs to top out on human IQ tests in late 2027, including when it comes to tests that require the AIs to look at the puzzles visually.

So I expect that we have 2 years where AIs are supplementing people more than supplanting.

Even after AIs exceed human IQ tests, which I’m sure they will, they’ll face many further hurdles before they have anything like human agency in the world. That’ll give us a number of additional years to figure things out.

So I think the data are in line with the views of middle-grounders like the excellent AI journalists Dwarkesh Patel and Timothy B. Lee.

We have more time to watch AI development and see where things stand, as they begin to approach the top of human intelligence.

Stay tuned!

TrackingAI.org updates as new data comes in, and we’ll soon be adding live-updating monthly charts like you just saw above.

Here are all four measures plotted together:

All this data comes from my site TrackingAI.org, which has tons of additional info and data on AI progress, well beyond what’s presented here. Credit to Hans Lorenzana, who I employ to do the coding on it.

2

sebjenseb

I Still Don't Believe The Average American College Student Has an IQ of 102.

Read more

2 years ago · 56 likes · 12 comments · Sebastian Jensen

ChatGPT’s latest paid model, o3, was released this week. It scored a stunning 136 on the Norway Mensa IQ test:

An IQ of 136 is in the top 1% for humans.

In contrast, here's how the leading AIs scored on the exact same test just 11 months ago, in May:

Incredible progress.

The one caveat is that the results are based on the Norway Mensa IQ test, which is public, and for which answers do exist online (though they are not particularly easy to find.)

To deal with that, last year I worked with Mensa member Jurij to create an offline-only test from scratch (so it'd be out of the reach of AI training data) and then I made the scoring of the two tests roughly equivalent, determining question difficulty by asking Maximum Truth readers to take questions from both quizzes.

Here's how AIs do on the quiz that’s outside of AI datasets:

Now o3 scores an IQ of 116, putting it in the top 15% of humans. The median Maximum Truth reader, for comparison, scored 104.

And here are the leading AIs on the exact same test 11 months ago, back when we were better at IQ tests than AIs:

Again, this is simply incredible progress for 11 months.

I sent this info to Tyler Cowen of Marginal Revolution, and he further noted, “of course those tests underweight the value of extreme breadth, which of course o3 has.”

That is also correct.

Here is how I would put the findings:

ChatGPT’s o3 model is something like a person with 116 IQ when reasoning from scratch, but furthermore is like such a person with the entire world’s knowledge in its memory.

When it comes to reasoning on questions with already established answers, o3 appears more like a person with a 136 IQ, but this is largely due to having the entire world’s knowledge in its memory, including the work of the smartest humans, even if it could not solve such questions from scratch.

Conclusions

One still hears people say, “AIs can regurgitate knowledge, but they can’t think.”

That’s wrong.

AIs don’t feel things, but they can think, in that they can solve never-before-seen problems by deducing complex patterns.

I have more to report, including about AI vision progress and how it will allow AIs to interact with the physical world, and AI politics, but that’ll be for a future post.

Just like the internet was mostly ignored in 1995, AI even now remains dramatically under-covered considering the impact it will soon have.

Deepseek has set some impressive performance benchmarks, and is also nearly tying ChatGPT-4o in the Chatbot arena, where users rank AI answers.

But how does Deepseek do on measures of pure intelligence? I’ve now run the IQ tests.

On publicly-available IQ tests, Deepseek ties or exceeds all free AIs, but is far behind OpenAI’s models for paid users

On my site TrackingAI.org, we regularly give AIs IQ tests. We describe the questions to the AIs verbally, as if one were giving the test to a blind person. Here’s how Deepseek does on the publicly-available Mensa Norway test:

Both of Deepseek’s models tie or exceed all other free AIs, including Claude, ChatGPT-4o, and Gemini Advanced.

But Deepseek remains significantly behind OpenAI’s paid subscription models, such as o1-preview, which was released in September, and which I reported at the time represented a “massive breakthrough” in AI intelligence. China’s entrepreneurs have not yet hit that level.

The above IQ test also potentially suffers from the problem that its questions and answers exist on the open internet, which could lead to AIs “cheating” by having the answers somewhere in their database.

So last year, I worked with a Mensa member named Jurij to create new questions that don’t exist anywhere online. Then I gave the questions he made to subscribers of this Substack, and “normed” the scoring on the new test so that it’s equally as hard for humans as the Mensa Norway quiz.

On IQ questions that only exist offline, Deepseek R1 beats all free models, but is behind OpenAI’s paid models

Here’s how Deepseek does on our quiz that doesn’t exist online:

We can see that Deepseek V3, which is optimized for “versatility”, got a mere 70 on the offline questions. So it was probably “cheating” on the Mensa Norway test, relying on answers from the internet.

But Deepseek R1 — which is optimized for “reasoning” — gets a very strong score on the offline-only test, around 90. It was beaten only by the OpenAI models that require a paid subscription to access (as in the first test.)

Example where Deepseek is less smart than OpenAI’s o1

Let’s look at question 24 on the Mensa Norway IQ test. Here’s the question:

It’s not an easy one. Maybe you can see the answer — or maybe you can’t.

Deepseek’s reasoning model fails, and it also doesn’t really explain its logic beyond “patterns … that share common angles”:

OpenAI’s ChatGPT-o1-pro (which costs $200/month) almost fails and uses incorrect logic, as well — but then it double-checks its logic, realizes it doesn’t hold up, and then finds the correct logic (highlighted in blue):

OpenAI’s o1-preview also nailed the question (it did so without first trying out a failed pattern.)

OpenAI’s paid model intelligence is really impressive. The paid OpenAI o1 models are at least as good at abstract thinking as the average human.

Deepseek was beaten by the OpenAI paid models on this question, and several others (in both the online and offline-only quizzes.) But let’s remember that Deepseek tied or beat all other models, so it is still very impressive.

Intelligence is different from user satisfaction

In Chatbot Arena, users are served two different AI answers side-by-side, and they rate which one is better. Based on those user satisfaction scores, we can see that model intelligence is not perfectly correlated with IQ performance. Specifically, two Gemini models and ChatGPT-4o (which is free up to a limited number of uses) are slightly beating Deepseek R1 in user satisfaction.

But Deepseek R1 slightly beats ChatGPT-o1-preview, despite the latter’s clear edge in IQ performance.

This shouldn’t be shocking, since humans are doing the rating; in the real world, I think we’d also often see that answer satisfaction doesn’t perfectly line up with IQ; a high school principal with an IQ of 110 would plausibly have much higher-rated answers than would a 140 IQ physics professor.

There is still a clear correlation between AI intelligence and user ratings — it’s just far from perfect.

The distinction is important: If you want an AI to develop new drugs, or create a post-scarcity society, or achieve the “the singularity”, then raw intelligence is critical. On the other hand, if you want to answer a question like “how do I jump-start a car?” then an IQ of 100 and a database of the whole internet works just fine.

Share

It’s good that the United States retains the lead in raw AI intelligence; we’ll see how long that lasts

Deepseek’s model is seen as extra impressive because it was trained using an older version of Nvidia chips — the newest ones were off limits to Deepseek due to US export controls designed to prevent China from winning on AI.

The Chinese entrepreneurs behind Deepseek also seem like well-meaning innovators — impressively, they open-sourced their project, meaning anyone can inspect and copy their code. The only major American player to do that is Meta’s Llama.

However, the Chinese government is one of the most authoritarian on Earth, and it’s plausible it will pressure China’s entrepreneurs to keep further AI developments more closely-held.

Extrapolating from American progress, it may not take long for Deepseek to reach the front of the pack. Deepseek is already better than ChatGPT-4, which was released 1.5 years before o1-preview. It’s also better than ChatGPT-4o, which was released just four months before o1-preview.

However, further progress may require Deepseek to find more shortcuts similar to the ones they found in order to train their current models. We’ll see if they can do that.

The increase in Chinese AI competition also increases the case for the US plunging forward with relatively unregulated AI, because while there are significant risks to developing AI at all, there are additional risks to an authoritarian state like China achieving super-intelligence first. That’s not necessarily definitive in a cost-benefit calculation on developing AI, but it is one relevant factor to consider.

American companies continue to innovate rapidly

Stay tuned for further posts detailing recent breakthroughs from OpenAI. Sign up here to make sure you get those posts!

You can find all the AI answers to the Mensa Norway questions on TrackingAI.org. Additionally, you can review the verbalized questions here, which also allows you to replicate my testing.

I’ve been following them all closely, because I run ElectionBettingOdds.com, so I’ll share what I’ve observed:

Trump first gained in the odds, then lost some of his lead

Here’s where the betting odds stand right now, along with the graph over the last month:

Looking at the start of the graph, bettors bid up Trump’s odds for a while, front-running the polling changes.

The VP debate was held on October 1, and from then through October 25, bettors raised Trump’s odds from about 47% to 57%.

Some were skeptical, and worried that “market manipulation” was at play, especially after it was revealed that a single French trader had bet $25 million on Trump on Polymarket.

But bettors’ predictions were soon confirmed by polling. See the graph for how bettors correctly anticipated Trump’s rise in polling-based models:

In the raw polling, one could see a shift as well. Trump at one point was leading in all the swing states, according to the RealClearPolitics average, which has the best historical track record among the different polling averages.

In last week, however, Trump has slipped slightly in the polls. RealClearPolitics now has him losing slightly in Michigan and Wisconsin.

Also, a poll in Iowa came out this weekend showing Trump losing Iowa — which had been considered a safe red state. That poll had a respected methodology, and it spooked markets this weekend, briefly sending Trump’s odds of winning Iowa down to below 80%, and sending his odds of being President down as much as 5 points.

NBC News correspondent Steve Kornacki explains how the new Iowa poll hints at a possible narrative where abortion policy is key:

The poll might still be way off — a different poll the same day had Trump up 10 points in Iowa — but the bettors think there’s about a 20 percent chance that the poll with Harris leading is right, based on the odds they give for Iowa:

The betting odds still give Trump an ultra-slight edge for becoming President — now essentially the same as what Nate Silver gives him.

The bettors are much more hesitant about predicting a Trump win than they were just a few days ago.

Subscribe now

Was there market manipulation?

Before Trump’s slide in the odds last week, one French trader’s $25 million on Trump had moved the markets slightly. Were his trades “manipulation”?

For comparison, over $2 billion total has been traded on the election, so $25 million is a little more than 1% of the total bet.

Looking at changes in individual markets always suggested that Trump’s odds were rising, regardless of any “whale” making trades:

PredictIt is a good “sanity check” to look at if you ever fear market manipulation, due to strict regulatory trading limits. Because of the way PredictIt was given government permission to operate way back in 2014, it alone among the major markets has to limit traders to $850 per contract. It must also limit the number of traders per race to 5,000.

As a result, PredictIt combines aspects of a poll which tracks the same people over time, and also a betting market.

Because of that, PredictIt is really difficult to manipulate. It’s also very hard to make real money doing arbitrage with it, so manipulating another market won’t automatically move PredictIt.

So whenever one worries about manipulation, you can compare PredictIt to the other markets, and see if it is moving similarly. If it is, then manipulation probably isn’t the only factor.

Note that the above does not mean PredictIt is a pure, unbiased, indicator, or that it’s any better than the other markets. PredictIt has its own biases and weakness. In particular, it has a specific user base, which may lean more towards political “nerds” then do the other markets. In general, PredictIt has leaned more Democrat for the whole election cycle.

But it’s good to have slightly different markets. And it’s a reason that ElectionBettingOdds.com’s average is useful; the average of truly different markets is likely to be more accurate than any single market, just as averaging polls is more accurate than any single one.

Trump’s odds, as I write this, show the following:

Note also that, while the above provides general evidence that Trump’s rise was not due to manipulation, the French trader did eventually talk with the press, and he made that clear as well:

Share

Polymarket’s volume appears to be inflated

In the screenshot above with the five markets, one stands out: Polymarket reports $3 billion traded.

But on Wednesday, Fortune revealed that as much as a third of Polymarket’s volume is due to fake “wash trading”:

In separate investigations completed by the blockchain firms Chaos Labs and Inca Digital and shared exclusively with Fortune, analysts found that Polymarket activity exhibited signs of wash trading, a form of market manipulation where shares are bought and sold, often simultaneously and repeatedly, to create a false impression of volume and activity.

They go on:

Chaos Labs found that wash trading constituted around one-third of trading volume on Polymarket’s presidential market

So while my odds screenshot above shows that Polymarket’s volume is $3 billion, the real number is probably $2 billion or less.

The good news is the wash trading doesn’t impact the odds, as prediction market pioneer Robin Hanson notes:

I had some suspicions about whether Polymarket’s volume was of the same quality as volume on other exchanges, and that’s one reason ElectionBettingOdds.com has not volume-weighted the odds this cycle. A further reason is related to what I noted above about PredictIt; a market’s predictive value can be due to not only volume, but also to having different trading rules.

More betting markets have entered the space

Thanks to a court ruling on October 2, betting on elections is now legal in the United States (for now, pending a final court ruling.) On October 4, Kalshi, the company which won the injunction against the government, listed their first Presidential betting markets.

On October 5th, I added Kalshi to the ElectionBettingOdds.com average.

Since then, more exchanges have come online with election trading, such as Interactive Brokers, and the even more famous Robinhood. Those are very streamlined markets, are fully legal, and have relatively low fees. They were launched too close to the election for me to add them to the ElectionBettingOdds.com average this cycle. But next time, they’ll be there.

Polymarket still has by far the most creative and liquid markets (notwithstanding the wash trading) but Americans aren’t supposed to trade there — anyone who does is using a VPN to access the site.

Manifold Markets is a formerly all-play-money site that recently also added an element of real-money trading in the form of “sweepstakes.”

In sum, the election trading space is more vibrant than ever, and there are plenty of legal sites where Americans can trade.

It’ll be interesting to see how the markets do on Tuesday. They have a good track record in the past.

The US election is coming up! We’re flooded with talking points, but I haven’t seen a single article going through the issues and earnestly making a case for a candidate.

Everyone thinks it’s “obvious” why their candidate is better. It’s not obvious!

So today, I’m simultaneously posting the strongest possible case for Harris, and also for Trump.

Here is the rational case for voting for Trump:

Trump does have flaws

Let’s get it out of the way: I am aware of Trump's drawbacks. He can be rude, thinks transactionally, has a colorful personal life, sometimes puts politics above factual accuracy, can be a sore loser, and makes things about himself.

Despite those things, there's a very strong case to vote for Trump this year:

Preserving US Institutions and Courts

While Trump is often portrayed as the threat to US democracy, it’s Kamala Harris who has endorsed eliminating the filibuster and packing the Supreme Court. That’s no idle threat, as every Democrat in the senate already voted to end the filibuster (besides the two who are retiring: Manchin and Sinema.)

Packing the Supreme Court would effectively end the rule of the Constitution, as judges would then be determined by whichever party were in power. It’s not hyperbolic to note that Venezuela started its descent into dictatorship and starvation-level poverty with court packing.

Trump, in contrast, appointed judges to the Supreme Court who read the plain language of the Constitution, and who aren’t afraid to cast tie-breaking votes that deviate from conservative consensus — whether that’s Gorsuch ruling in favor of enforcing ancient treaties with Native Americans, or Amy Comey Barrett ruling that Arizona can’t revoke the voter registrations of people who didn’t provide proof of citizenship, or Brett Kavanaugh ruling against Republican redistricting plans.

Trump’s judicial appointments have been excellent in general, and contrary to stereotype, an academic study finds that, “across three different measures: opinion production, influence (measured by citations), and independence … Trump judges outperform other judges.”

As long as the Supreme Court is made up of such independent-thinking judges, the risk of any authoritarian takeover or successful “coup” is extremely low. The biggest risk to Constitutional Democracy is the destruction of the court itself, which Democrats increasingly advocate for.

Fewer Regulations

Trump, and Republicans, are instinctively anti-regulation. For example, Trump:

• Won’t ban prediction markets, as Biden’s CFTC commissioners are in the process of doing

• Won’t throttle cryptocurrencies

• Won’t limit fracking and energy development

• Won’t add costly, needless airline and car regulations

Harris’s bureaucrats would restrict our freedom to innovate in a hundred other regulatory ways as well, which Trump’s people would avoid.

Avoiding overregulation is a big deal. An academic study found that, “Had regulation been held constant at levels observed in 1980, our model predicts that the economy would have been nearly 25 percent larger by 2012 … regulatory growth since 1980 cost GDP $4 trillion in 2012, or about $13,000 per [person].”)

Elon Musk has also become a target of overregulation. California authorities recently shot down launch permits for Musk’s satellites, with one regulator (Gretchen Newsom) showing malice by blasting Musk’s politics at the regulatory meeting: “Elon Musk is hopping about the country, spewing and tweeting political falsehoods and attacking FEMA.”

Musk’s innovations are so important. He has already provided worldwide internet by launching more satellites than all governments combined, and it would be amazing if he could get us to Mars next. He’s already enabled people to play Civilization VI using only their brains, and it would be amazing if that can progress into helping people with mental health issues.

Protecting Musk’s projects from government interference is by itself a small reason to vote for Trump, and there are millions of other entrepreneurs who will also do cool things if not overly restrained.

Preserving free speech

Mark Zuckerberg recently spoke out about how the Biden administration pressured him and Facebook to censor Covid content, leading to absurd decisions like censoring the "lab leak" theory, and even censoring experts who argued against masks in schools.

Kamala Harris endorsed censorship wholeheartedly in 2019, saying about social media companies, “if you profit off of hate, if you act as a megaphone for misinformation … if you don't police your platforms, we are going to hold you accountable.”

This year, Harris's running mate, Tim Walz, said, “there’s no guarantee to free speech on misinformation or hate speech.”

That is dead wrong when it comes to the US Constitution, which protects those things because laws against so-called “misinformation” and “hate” would allow government to censor any speech it dislikes.

The United Kingdom shows what unrestrained progressive leadership looks like. The UK arrests thousands of people every year for offensive social media posts, and recently jailed someone for TWO YEARS for posting things like “it’s OK to be white.”

We owe it to Trump's 2016 victory that free speech is not the top concern right now, because (barring court-packing) this court will uphold free speech. If the court were instead 6-to-3 left-leaning, then reinterpreting the First Amendment to exclude "hate speech" and “misinformation” would be a hot legal topic. A Harris Presidency would move us closer towards that possibility.

Relative Focus on Merit

When President Biden got the opportunity to nominate a Supreme Court justice, he promised at the very start of his selection process: “that person will be the first Black woman.”

When President Biden picked his Vice President, he promised, well before picking anyone, “I commit that I will in fact pick a woman to be Vice President.”

Such explicit criteria for a particular role are not normally even legal in the private sector. It’s no way to run a government, either.

Trump will govern in a more colorblind and genderblind way. So if Vivek Ramaswamy ends up in Trump’s administration (as I hope he does) it won’t be because he’s Indian. It’ll be because he’s brilliant, and ideologically-aligned with Trump.

Beyond “DEI” concerns, compare the massive quality difference between Trump’s new top advisors and surrogates with Harris’s.

Elon Musk takes the cake, but the FT reports that Silicon Vally executives’ donations as a whole have turned Republican! For the first time since 2000:

The list of Silicon Valley executives now backing Trump include David Sachs, Joe Lonsdale, the Winklevoss twins, Andreessen Horowitz, Chamath Palihapitiya, and more. JD Vance also worked in Silicon Valley, investing in startups.

I can’t think of anyone I’d rather have pulling the strings of government than such innovators.

Contrast that with Harris’s circles, which are heavy on DC political hacks, over-credentialed ladder-climbers, bureaucrats, and activists. No thanks.

The same difference also shows up in the candidates’ own life experiences. Harris and Walz have approximately zero private sector experience, in stark contrast to Trump and Vance:

No wonder Trump and Vance are better when it comes to regulation and the economy.

Better on Wokeness / Trans Issues

Kamala Harris bragged to the ACLU in 2019, “I pushed the California Department of Corrections and Rehabilitation to provide gender transition surgery to state inmates.” She also said it should be given to incarcerated illegal aliens.

Should you and I, as taxpayers, really pay for that?

Joe Biden's pick for Supreme Court, when asked to define the word "woman", also famously responded that she couldn't, because "I’m not a biologist.”

I understand where people like her come from — I don't want trans people to feel bad either. There's no need to constantly remind trans individuals about biological sex.

At the same time, I expect serious people like the President of the United States, or a Supreme Court Justice, to be able to say obvious truths.

Otherwise we end up with absurdities like US prisons, and the military, paying for sex change operations, putting biologically male prisoners in women's prisons, and women getting beat up in sports by people with XY chromosomes.

It’s a case of "pathological empathy" that, at the highest levels of government and academia, people have decided to play along with gender dysphoria. Denying truth never works out well. Meanwhile in everyday life, people can still be encouraged to be nice and supportive of those who live in a very different way.

Trump walks the balance between truth and empathy on this better than Democrats, by not denying basic truths, while also not gratuitously making fun of trans people, and while welcoming them and other LGBT Americans into his coalition.

Intellectual Tolerance

Democrats, incredibly to me, often try to get people fired for stating facts like the above. Republicans have somehow become the party of relative intellectual tolerance.

When RFK Jr. decided that he had no path to victory this summer, he reached out to both the Trump and Harris camps. RFK Jr. says that Harris “declined even to speak with me.” He ultimately endorsed Trump.

I disagree with RFK Jr. about vaccines, energy policy, and more, but I despise the culture that says he's "beyond the pale" just for coming to different conclusions. All Americans should be able to talk with one another.

Creativity

During Trump's first term, he suggested that the US should buy Greenland. That’s creative. I’d never thought of it. It’s a cool idea, makes sense geopolitically and economically, and while Denmark likes to claim that Greenland is not for sale — maybe they’d change their mind for a trillion or two.1 Buying Alaska also seemed weird in 1867, when people widely ridiculed the purchase.

JD Vance shows every bit as much intellectual creativity. For example, he floated the idea of giving votes to children under age 18, and assigning those votes to their parents. Again, it’s something I hadn’t thought of. The policy would culturally encourage people to have children, and it’d give more control to those who have a greater stake in the future of the country.

Such ideas are speculative. They need more debate.

But how refreshing is it to have leaders who are creative enough to come up with new ideas? As opposed to regurgitating the same consultant-created talking-point slop that we’ve been force-fed for decades?

Crime

The President doesn’t have much authority to intervene when it comes to local crime. But at least, when there were mass riots in 2020, Trump shouted that someone should do something rather than tweeting links for bail funds while rioters were burning cities down:

The bail fund that Harris promoted at the height of the riots ended up bailing out a man on his 8th count of indecent exposure, who went on to murder someone once he was out on bail. It also bailed out a man accused of domestic assault, who also went on to murder. In other words, the charity may not have been the most effective altruism.

At the same time, Tim Walz was governor of Minnesota. While hundreds of local businesses were burning in his city, his wife was enjoying the poetic nature of the scene: “I could smell the burning tires,” she said. “I kept the windows open as long as I could because I felt like that was such a touchstone of what was happening.”

Self-congratulatory masochism, instead of focusing on the people whose lives’ work were going up in smoke.

Meanwhile, Walz’s daughter was tweeting to protestors and rioters, “the National Guard WILL NOT be present tonight.”

Walz’s family aren’t running for President. But they give us a good sense of what’s within the “Overton window” in the Walz household.

One candidate, and one party, is dramatically more sane on the issue of crime.

Immigration

I support more high-skilled immigration. But the situation on the southern border has been a disaster. It was almost entirely caused by the Biden-Harris administration, on day one, revoking all of Trump’s executive orders and international deals that were keeping the border secure:

Harris, instead of saying “we made a mistake,” (or, more honestly, “my administration is staffed with ideologues who don’t believe in borders”,) disingenuously blames Congress for not passing immigration bills, which she knows contain “poison pills.” The bill proposed in 2021 would have provided amnesty to 11 million people, and the one this year would have codified keeping the border largely open.

Picking the future citizens of the United States is a serious and irreversible policy issue, and so it’s particularly important to decide the question openly and democratically, with broad buy-in from Americans. Instead, we get the Biden-Harris administration ignoring the rule of law by letting millions come and stay indefinitely.

It may be true that immigrants, on average, don’t commit more crime than Americans. But should that be the bar? Did America really need to let in the Venezuelan gang members who have forcibly taken over an entire apartment complex in Colorado? Or the migrants who are causing 75% of all arrests in Midtown Manhattan?

We can do better when it comes to picking our future fellow citizens. Trump kept illegal crossings low, and he keeps an open mind about legal immigration, earlier this year even saying, “you graduate from a [US] college, I think you should get automatically, as part of your diploma, a green card to be able to stay in this country.”

Foreign policy

Trump can sometimes seem unorthodox on the world stage, maybe even a bit crazy. But there’s a method to his madness.

How do I know? Look at this off-camera exchange Trump had with Wall Street Journal editor Paul Gigot, just days ago:

Mr. Gigot: “Would you use military force against a blockade on Taiwan?”

Mr. Trump: “I wouldn’t have to, because he [Xi Jinping] respects me and he knows I’m f— crazy.”

Trump has practiced negotiation more than you or I, and his quote suggests that he well understands the benefits to being perceived internationally as dangerous and volatile.

In contrast, we’ve seen the fruits of the Biden-Harris strategy of cautious containment, where they give just enough aid to anyone fighting America’s adversaries to keep things in a bloody stalemate.

The results from Trump’s last term speak for themselves: Foreign bad actors were scared to act out, and there were far fewer wars in the world.

Abortion

Trump's position on abortion is the following:

• Roe v. Wade was wrong (it was)2

• Voters in states should decide; he would veto any national abortion ban

• Total bans go too far; at least have exceptions for rape, incest, and life of mother

• Allowing abortion until birth for any reason, as many blue states do, goes too far

Most people don’t know that Trump has taken many concrete steps to moderate his entire party on abortion:

1) Trump successfully pressured Alabama to almost immediately reverse a law that courts had interpreted as preventing IVF fertility treatments.

2) Trump successfully pressured the Republican Party to remove a national abortion ban from its platform, which had been in there for 40 years.

3) Trump has even proposed requiring insurers to cover IVF. Government mandates like that are non-libertarian, but might at least help with declining birth rates.

In contrast to Trump’s moderate and nuanced approach to abortion, from Harris we get absolutism. Seven blue states now allow abortion all the way up to birth, for any reason at all — a policy more extreme than that of any other country in the world.

Transparency and consistency

Harris now claims she doesn't want to ban fracking,3 a complete reversal of her 2019 quote that, “There’s no question I’m in favor of banning fracking.”

Harris now claims she doesn’t want to force Americans to sell their guns to the government, a total reversal of her 2019 quote that, “we have to have a buyback program [for assault weapons]. I support a mandatory buyback program.”

I have never seen Harris, ever, explain why she changed her mind on so many of her positions. If you listen carefully, she doesn’t even admit she changed them. That’s not honest, and it gives me no confidence that she has moderated her views.4

By contrast, when JD Vance is asked about his changes in views, he just says he changed his mind, and then gives plausible reasons why. Is that too much to ask?

Outcomes are hard to predict

The impact of any President is hard to predict in advance, because there are so many factors in play. For example, will the “woke left” have a resurgence if Trump wins? I doubt it; I think they’re exhausted. But nobody can know for sure.

In general, we should avoid overthinking about second-order effects and just ask: Which Presidential candidate supports sane economic policy, sane judges, and sane immigration rules? Which candidate is unafraid to speak basic truths? Which candidate is more creative and open-minded? Which candidate has the better track record, both in terms of freeing entrepreneurs from taxes and regulations, and in terms of keeping the world at peace?

The answer to all of those questions is Donald J. Trump.

Disagreement is Reasonable

I think I've presented a compelling case here, but you may have different priorities and values, and may disagree. That’s fine. I hope my arguments here at least made you think. Perhaps you also now better understand why many non-crazy people will vote for Trump. Maybe you should, too!

For the counterpart essay, please see “The Rational Case for Harris” at MaximumTruth.org

This is a guest post by a friend. I am posting my friend’s essay unedited and as-is, in terms of editorial content.

Introduction

In the history of US Presidential elections, the rational decision for voters has rarely been clearer: Kamala Harris is the correct choice for President. For moderates, and for those few remaining undecided voters, here are 4 of the best arguments for Kamala Harris’ presidency.

Harris Is Better For The Economy

The economy is doing well now, and will continue to do well under democratic leadership. The evidence for this can be seen in the unemployment data. In the third quarter of 2023, GDP grew by 4.4%; in the fourth quarter of 2023, GDP grew by 3.2%; in the first quarter of 2024, GDP grew by 1.6%; and in the second quarter of 2024, GDP grew by 3.0% (Bureau of Economic Analysis, Annual Update, September 26, 2024). Four quarters of growth, combined with similarly strong post-pandemic growth in general since mid-2022, has put the US economy ahead of every other advanced economy in the world since Biden took office. From February 2022 to April 2024, Biden oversaw the longest period of sub-4% unemployment since the 1960s.

Furthermore, despite the widespread misperception that recent revisions in job growth estimates indicate a weaker economy, “total nonfarm payroll employment increased by 142,000 in August, and the unemployment rate changed little at 4.2 percent.” (Bureau of Labor Statistics, September 6, 2024). September 2024 job growth was similarly positive.

Inflation is also down to around 3.7% from its pandemic high of around 6.6%, indicating that Democratic economic policy has been effective in curbing inflation while avoiding pessimistic predictions of a post pandemic recession (Source). Conservative commentators sometimes argue against these numbers in conspiratorial ways, saying that the various metrics used by the Fed and other agencies to measure inflation, unemployment, and GDP growth do not represent the ‘true’ or ‘real’ state of the economy. But investors are not falling for conservative counter-messaging: in recent weeks, the stock market hit new highs, including the DOW and Tech sector stocks, and IPOs are seeing high valuations.  

The economy will continue to do well in the future under Democratic leadership. First, given the stellar post-pandemic economic recovery, if Harris simply continued Biden’s policies, the economy will continue to grow and prosper. But Harris has a plan to build on Biden’s wins, which she elucidated in an 82-page dossier this month. Her plan includes policies to cut taxes for working class Americans, reduce the cost of groceries, lower healthcare costs, including the cost of prescription drugs, and lower energy costs. Her plan will address average Americans’ concerns regarding the rising cost of living in America. Many Americans have the mistaken perception that the economy is not doing well in part because the cost of groceries, housing, consumer goods and services, and energy are all higher than 4 years ago, and wage growth has not offset these high costs. Harris has a plan to fix that so that working class Americans can experience the wider economic prosperity that upper class business owners like Trump are currently enjoying.

Harris will be, not just good, but better for the economy than Trump. Trump does not have a clear plan to contrast the Harris plan against. He has concepts of a plan for healthcare, which could become part of a comprehensive economic plan he has been promising to reveal since 2016, but as of this writing, Trump has not produced anything as concrete as the Harris campaign regarding economic policy. At most, one could reasonably surmise from his characteristically vague pronouncements that Trump intends to enact disastrous tax cuts for big businesses and wealthy Americans such as himself, just as he did when he was president, and introduce tariffs that will hurt working class consumers. As Hannah Kling, assistant professor of data analytics and economics at Belmont Abbey College, told the National Catholic Register recently in her bipartisan, balanced criticism, “that tariffs and price-gouging laws are extremely popular among the general public while all economists know they only hurt the general public really breaks my heart.”

The cost of tariffs on foreign goods will be passed on to consumers. Raising costs for consumers will spur a manufacturing renaissance by creating demand for domestically produced goods, if Trump is to be believed. But as Dr. Kling said, all economists know this thinking is wrong. Trump’s current promise to bring back manufacturing jobs is very similar to the empty promises he made during his time as President, by the end of which the Rust Belt states had lost hundreds of thousands of manufacturing jobs, a trend that only accelerated in the wake of Trump’s failed response to the pandemic. To be clear, the damage that Trump inflicted on the Rust Belt cannot be attributed to the pandemic because almost 1,800 factories disappeared during the early years of Trump's administration between 2016 and 2018; but Trump’s slow response to the pandemic did exacerbate the manufacturing recession he presided over.

Bringing it all together, then, the argument is clear. The economy is doing well now, and it will continue to do well in the future under Harris and Democratic leadership. Trump’s economic agenda failed in the past, and Trump’s tax and tariff policies are likely to fail again in the future. The conclusion a rational thinker will draw from this evidence is that Harris is better for the economy than Trump.

Harris Is Better For Foreign Policy

Harris has a better temperament for wielding power. She is cool under pressure and respected by world leaders. On the war in Gaza, she has expressed the appropriate amount of support for Israel and she met with Netanyahu, while also expressing concern for civilians in Gaza. Harris is also well positioned for a post-Netanyahu Israel, having voiced strong support for peace between Israel and its adversaries, and having a husband who is Jewish, which gives her unique access to Jewish lived experiences. Harris also has close ties with leadership in Japan, Mexico, Jordan, the Philippines, and Ukraine.

Beyond her positive relationships with American allies, Harris will be a far more reasonable and level headed thinker than Trump if she is ever faced with a 5 minute countdown to launch a nuclear response. Americans would be able to rest easy with Harris carrying the nuclear codes. Her mind is clear and character is stronger than her opponents. Moreover, the single major criticism of her foreign policy record is only applicable if one buys into the fallacy of blame by proximity: she was reportedly the last in the room with Biden during the disastrous withdrawal from Afghanistan. Trump, it should be noted, was not able to pull off a withdrawal from Afghanistan while president even after promising to do so, asking his advisers if it was possible to do so, and undercutting the Afghani Government by negotiating a deal directly with the Taliban in an attempt to do so. He even planned to invite the Taliban to Camp David just before September 11th as a reward for the deal, but flip flopped in response to public outcry.

In the end, it is character that determines how well a president will lead, according to General Stanley McCrystal. When it comes to character, there are few Presidents in modern American history who have been as selfish, vindictive, and avaricious. Dangerously, Trump is cunning in his moral turpitude. He is a convicted rapist and serial philanderer, a scam artist who sells whatever he can, and a Judas who feigns interest in Christianity for his own gain; yet he has so far evaded any political accountability for superlative moral failings. He deserves to lose at the ballot box for those reasons alone.

Importantly for foreign relations, when provoked or flattered, he is easily manipulated. During the only debate with Harris that Trump has so far agreed to, in response to Harris’ goading remark that Trump is not respected by world leaders, a fact the truth of which Trump does not enjoy, Trump was flustered. He then resorted to bragging that the authoritarian leader of Hungary, Victor Orban, who has been president of Hungary since 2002 after undemocratically rewriting the Hungarian constitution in 2011, is a strong supporter of his candidacy, as if that were a good thing. This answer at once betrays both how much Trump cares about what other people think of him and how easily he can be manipulated through flattery (by Orban) and mockery (by Harris). That exchange was a small example of the much deeper problem that would face America if Trump is elected for another term: he is a narcissist, and every narcissist has roughly similar psychological tendencies. They need endless positive attention, and they will attack or avoid anyone who does not provide obsequious adulation. That is a weakness America’s enemies often exploit.

To underscore this point: Harris was able to get under his skin so easily during the debate because she understands that Trump is a narcissist, as do America’s enemies. As is true of every narcissist, Trump is extremely image conscious. Harris has prosecuted many such people and knows them well. Insult their pride, and they will show their vindictiveness, as Trump did when he turned on Netanyahu after Netanyahu refrained from joining in Trump’s election denialism and, instead, congratulated Biden for winning the 2020 election. Reward their greed, and they will do the bidding of whomever pays them, such as when Trump allegedly received a $10 million cash bribe from the president of Egypt in return for “releasing $195 million in military aid that was previously withheld because of human rights abuses committed by the Egyptian government, and he later released $1.2 billion more in military assistance,” according to the House Committee on Oversight and Accountability.

Putin clearly understands Trump’s narcissism as well as Harris does. Putin surrounds himself with oligarchs like Trump, and, in the same way that Putin expects loyalty from his oligarchs, Putin evidently enjoys Trump’s loyalty, rewarding Trump with an exclusive relationship that has included many secretive, closed door, one-on-one meetings. Records of what was discussed between Trump and Putin, one of America’s greatest adversaries, do not exist, and Trump went to great lengths to conceal the contents of these meetings even from his own administration. As a result, we may never know why Trump loved meeting with Putin so frequently, but whatever the reason, Putin has used Trump’s narcissistic tendencies to manipulate him into advancing Russia’s anti-American agenda by, for instance, advocating that Ukraine capitulate to Russia.

Beyond Trump’s weakness of character, he was also a horrible administrator as President. Trump displayed no aptitude for fortifying the organ of governance that the US relies on to manage foreign relations: the State Department. Trump left the department in shambles, through high rates of turnover, including one quarter of senior officials, and unfilled vacancies in nearly a dozen Assistant Secretary or Under Secretary posts, as well as extremely low morale and a wave of employee departures (the State Department lost about 12 percent of its foreign affairs specialists in the first 8 months of Trump’s administration.) Given the severe damage Trump inflicted on the department during his term, it is not clear that the State Department, and American foreign policy in general, would survive in an effective form, let alone bring stability to international relations, under another incompetent Trump administration.

More likely, Trump will sow chaos and disorder abroad, as he did when he randomly dropped a MOAB on Afghanistan for no strategic advantage, pulled out of the World Health Organization in the midst of the pandemic, canceled the Nuclear deal with Iran, and then assassinated a high level Iranian general, thereby setting US-Iran relations back by a decade and radically increasing tension across the Middle East. Trump will lower America’s standing in the world by damaging trust with US allies and capitulating to our enemies. Harris is the only responsible choice for improving American foreign relations.

Harris Is Better For Women’s Rights

Trump famously said that he liked to “grab [women] by the pussy,” which is a form of sexual assault, and later, he was convicted of sexually abusing (ie, raping) E. Jean Carrol by a jury in New York. Allegations of further misconduct swirl around Trump, with onlookers noting that Trump was very close with notorious rapist Jeffrey Epstein, as well as noting a 2016 case regarding allegations that Trump raped Katie Johnson when she was 13 years old. Trump regularly demeans women, having called them “fat pigs,” “dogs,” “slobs,” and “disgusting animals” at various times. Trump has bullied women for years this way, for example, saying of one contestant on the Celebrity Apprentice that “it would be a pretty picture to see her on her knees.” Trump has similarly insinuated that Harris is only successful because she performs fellacio on men. More recently, during the presidential debate, Trump continued to demean Harris by accusing her of being sexually promiscuous, saying “All I can say is I read where she was not Black, that she put out.”

As the evidence above demonstrates, Trump is a misogynist. He has no empathy or respect for women’s rights. He treats women’s healthcare with shocking glibness. Trump recently declared himself the Father of IVF despite having learned what IVF is only days beforehand. That kind of bizarre pronouncement exemplifies the kind of carelessness Trump brings to conversations about women’s rights. He is directly responsible for the repeal of Roe vs. Wade, which has given conservatives the power to use the government to control what women do with their bodies in healthcare settings. Here is what Melania Trump had to say on this matter:

“It is imperative to guarantee that women have autonomy in deciding their preference of having children, based on their own convictions, free from any intervention or pressure from the government,” wrote Melania Trump in her recent memoir, and continuing, “Why should anyone other than the woman herself have the power to determine what she does with her own body? A woman’s fundamental right of individual liberty, to her own life, grants her the authority to terminate her pregnancy if she wishes. Restricting a woman’s right to choose whether to terminate an unwanted pregnancy is the same as denying her control over her own body. I have carried this belief with me throughout my entire adult life.”

Evidently, Trump’s own wife does not believe that he is a good advocate for women’s rights.

Harris, by contrast, deeply respects other women, and as a woman herself, she understands the importance of reproductive rights from a personal perspective. Harris wants to keep the government out of women’s healthcare decisions. She will support a woman’s right to choose what she does with her body. Harris will pass legislation to make it easier for women to receive the care they need and deserve from healthcare providers. Considering the above evidence, the following statement is obvious, but for completeness and clarity: Harris is better for women’s rights.

Harris Is Better For Democracy

The final reason to vote for Harris is the clear threat that Trump poses to American democracy. This point follows naturally from the contrast between the two candidates. On the one hand, Harris is a star prosecutor, the former AG of California, and has a deep understanding of Constitutional law. On the other hand, Trump has little understanding or respect for the constitution. He wishes to “terminate” this sacred document of American democracy and rewrite it, just like his pal Victor Orban did when Orban rewrote the Hungarian constitution. Moreover, Trump blatantly attempted to take power and effectively end Democracy in the US following his election loss to Biden in 2020. Trump did this in at least two ways that we know of: first, by leading an insurrection on January 6th, 2021, and second, by attempting to tamper with the election he lost. Together, these twin efforts amounted to an attempted Coup d’État.

Throughout history, a standard way in which authoritarians take power is by using the Executive office they hold to attack the Legislature that represents the people. The Russians have a long history of engaging in this style of power grab. Under Lenin, the Bolsheviks stormed the Winter Palace and overthrew the fledgling Russian Republic in 1917. In 1991, Yeltzin shelled the Russian Parliament. Beyond Russia, many other examples exist. The Nazi Party set fire to the Reichstag in order to precipitate Hitler’s power grab in 1933. Napoleon dissolved the French legislature during the Coup of 18 Brumaire in 1799. Pinochet overthrew the Chilean government and dissolved the Chilean Congress in 1973. The authoritarian Fujimori dissolved the national legislature when he became dictator of Peru in 1992. Recently, in 2012, Morsi tried a similar move in Egypt by attempting to sideline the legislature to remain in power.

The pattern is clear. A tried and true method of overthrowing democracy is for the Executive to negate the Legislative. When Trump sent his supporters to storm the capitol building and prevent the US Congress from certifying the 2020 election results on January 6th, 2021, his failed coup attempt was added to the above list of historical attacks by executive power on legislative power. If Trump is elected again, he will likely take what he learned about his failed coup attempt in 2021 and implement a more effective attempt in 2028. In July, he said to his followers at a rally, "in four years, you don't have to vote again. We'll have it fixed so good, you're not gonna have to vote."

In addition to leading an insurrection against the Legislative branch of the US government, Trump attempted to end Democracy in the USA by tampering with the results of the 2020 election at the state level. There is not enough room to go into all of the details of the anti-democratic actions Trump took to alter the outcome of the 2020 election, from his request that the Secretary of State of Georgia “...find, uh, 11,780 votes...” for which Trump was impeached a second time, to his fake electors plot. But suffice it to say, Trump, cunning as he is, will do more damage to the fidelity of our election system in the future if he loses the 2024 election, and if he wins, voters would do well to take him at face value when he promised that Americans will not vote again in 2028.

The message is clear. The pattern is clear. A vote for Harris is a vote to preserve democracy, while a vote for Trump is a vote for the dissolution of democracy.

Note: The counterpart to this essay, The Rational Case for Trump, is available on the homepage at maximumtruth.org.

Are the critics right?

Before researching this post, I assumed that they were right about a single vote being irrelevant for Presidential elections. However, I still thought voting could be justified on “cultural/collective-action” grounds: Sure, your one vote will never matter, but if smart/ethical/pro-freedom people create a general cultural norm of voting, that can be enough to sway elections.

I’ve now gone through the literature on this subject. It turns out that, when one correctly calculates the chances of one vote flipping an election, one gets very high dollar values for a single vote.

Specifically, I estimate that a single correct vote for President in any of the seven key swing states is worth $10,000 or more, using low-end estimates. A correct vote in semi-competitive states is worth between $500 and $10,000 (that category includes states like Florida, Texas, Alaska, Minnesota, New Mexico, and New Hampshire.) It’s only “irrational” to vote for President in true blue/red states like Illinois or Wyoming, where the outcome is essentially certain.

Below, I explain in detail how I got those estimates.

Subscribe now

How likely is a single vote to flip an election?

I enjoyed diving through the academic literature on this.

Zach Barnett, an Assistant Professor of Philosophy at Notre Dame, has a paper showing that previous research on the issue used the wrong statistical distributions in estimating the chances of a vote tipping an election.

Barnett’s logic for his calculation method is elegant and clear, and he illustrates it with a hypothetical election between “Donald” and “Daisy” that has exactly 1,000,001 voters. He lists all the possible vote combinations:

1-in-a-million is the hard, proven, lower bound. But we also know that vote combinations in the middle of the distribution are dramatically more likely than at the edges.

Barnett goes on to show that the chance of a single vote flipping an election is at least 5x higher for the most competitive races (below, “d” means the chance your vote in decisive, and “N” means the number of voters):

So what does “4 / N”, etc, mean in the real world? Let’s zoom in on Pennsylvania, often the most swing of all swing states. Currently, Kamala Harris has a 40% chance there:

In Pennsylvania in 2020, there were 6,840,276 people who voted for Trump or Biden. If the same number of people vote this year, then the chance that one additional vote in the state will determine the election is at least 4/N, or 4/6,840,276.

Simplifying, the chance is 1 in 1,710,069.

So we get a real-world lower-bound estimate that a single vote in Pennsylvania has roughly a 1-in-2-million chance of determining the President.1 If the Pennsylvania race tightens again to being 50%-50%, then it will be closer to 1-in-a-million.

In other words, if you vote in Pennsylvania, you’re essentially buying a lottery ticket that you’ll get to decide the election. Your odds are actually a lot better than for winning the popular “Mega Millions” lottery (1-in-300-million.)

That result is also plausible looking at history. While a 1-in-2-million chance has not yet occurred in the short history of American Presidential elections, when surveying America’s tens of thousands of local elections, we DO see plenty of races being decided by a coin toss or drawing lots from a hat due to being an exact tie, where a single vote would have been determinative. There are tens of thousands of local races in a major election year, and local races often have tens of thousands of voters, so consider that some real-world confirmation of Barnett’s math proofs.

How valuable is an election outcome?

In order to estimate the value of a vote, we need to have a sense of the value of an election.

There are no academic estimates on that.2 People who have written on this instead use guesstimates. Philosopher Will MacAskill guessed it this way in 2015 (in the following quote, I’ve updated his numbers for 2024): “total spending of the US government is $3.5 $6.1 trillion per year … if that money is spent 2.5 percent more effectively, then the benefits amount to $1,000 $1,753 per person… $314 $547 billion…”

Guessing is a better approach than having no estimate at all, but, below, I’ll look at stock market changes to get some empirically-grounded sense of how much a vote could be worth.

My approach below was to estimate the societal value of electing the better President by examining the absolute change in the stock market (S&P 500) on the day following Presidential elections, going back twenty years to 2004.

Here’s the change the day after election day, for each election:

We could refine those estimates a bit. The refinements don’t make a difference to the general magnitude of the value of a vote, so I won’t bore you with the details, except in a footnote.3 But basically, I do two things: I remove the median daily stock swing, and adjust for how much prediction markets had already priced in candidates’ wins.

The chart of whether markets already expected wins is interesting in its own right:

After adjusting for that, the median change was from the 2020 election, where Biden’s 20% increase in odds from election night appeared to be worth about 1.71% of the stock market, indicating that his election as a whole was worth about 6.57%.

The lowest-impact election was 2016, in which Trump’s win over Clinton was valued at 1.39% of the market.

So that suggests that the value of a vote ranges from 1.39% to 6.57% of the S&P 500.

In 2024, with the S&P worth $55 trillion, that works out to $767 billion – 3.61 trillion.

There are lots of reasons why my measure is imperfect. The goal isn’t to get a perfect measurement, it’s to get a number that’s in the right order of magnitude. If it’s more sound than MacAskill’s guess, then we’re making progress. In fact, my lower-bound estimate turned out to be not that far from his guess (though I looked his up afterward doing the calculations.)

One further issue is that we’re looking at the change in the stock markets caused by all elections on election day, including House/Senate/Governors, etc.

It turns out that total donations to Presidents are only a quarter of all donations to federal candidates, and only 23.67% if one also considers spending on governor’s races.4

So let’s estimate that the Presidency is worth 23.67% of all elections on election day (that feels low, but that’s what donation stats suggest, and they’re a “market” in a sense.) Then our new estimated range for the Presidential election goes down to $182 billion -  $852 billion.

That estimate feels low. It amounts to about a mere 3% to 15% of one year of government spending. It’s just 0.7% – 3.4% of one year’s economic output.

It seems especially low when one considers that any effects of Presidential policy are compounding, and continue far into the future. For example, the 2017 Trump tax cuts still impact the economy 7 years later. Obamacare still affects us 15 years after enactment.

The stock market does include those indefinite effects (with discounting) in its $182 billion - $852 billion implicit estimate.

I should note that social issues and foreign policy may each be just as important as economic policy, but they can’t be quantified here. So this is, in every sense, a lower-bound estimate on the value of an election.

But at least we now have a ballpark number for that.

Let’s now see what happens when we combine it with the 1-in-2-million odds we calculated earlier.

The Enormous Value of A Correct Vote in a Swing State

First, we’ll take Nate Silver’s model’s estimate (paywalled)5 that Pennsylvania has a 29% chance of tipping an election. Then, the expected value of your vote is:

[Chance of state determining Presidential election] * [Chance of voter determining state] * [Value of best candidate winning]

So:

Low-bound estimate: .29 * (1 / 1,710,069) *  $182 billion

Higher estimate: .29 * (1 / 1,710,069) * $852 billion

Doing the math:

Low estimate:  $30,864

High estimate: $144,485

Wow! The first time I crunched those numbers, I was shocked. $30k per vote, as the low-bound?

But I keep checking over the math, and it appears to be right. Also, while this estimate is empirically-backed, if you prefer any other plausible guesstimate about the value of electing the better President — you’ll still get massive values.

Looking beyond Pennsylvania, here are the other states:

There is, however, one possible factor we haven’t included yet:

Problem: You don’t know who the better candidate is!

The above assumes that you, as a voter, know which candidate would be better for the country.

However, as philosopher Jason Brennan points out in this paper (paywalled),  if you vote for the wrong candidate, you cost society the same amount of money. Brennan notes:

“Remember: if a vote [for the right candidate] is like donating $3,000 to a charity, a vote the other way is like destroying $3,000 of their funds.”

So let’s say you’re barely better than the median voter at selecting the right candidate. Let’s say you’re only 60% sure that you picked the better one.

Then the value of your vote in Pennsylvania would shrink to: (60% - 40% = 20%) of the original estimate.

So if we assume you should only be 60% confident in your vote, then a Pennsylvania vote is actually worth $6,173 to $28,897 (instead of $30,864 to $144,485.)

That’s still worth a lot. Even you’re deeply conflicted, but have a gut sense who will be better for the country (60% sure) it’s still well worth your time to vote.

Here’s how the map changes, with that assumption:

Now, most of us are more than 60% sure. Maybe we’re 95%, or 100% sure. But should we be that sure?

History gives us lots of room for humility.

Why we should be humble about whether we know which candidate would be best

The famous biography of Sam Bankman-Fried, “Going Infinite”, documents how, when Bankman-Fried was a trader at Jane Street in 2016, he brilliantly modeled the election night results, and called Trump’s win before anyone else. He then used that information to short-sell the stock market, as he assumed that a Trump win would be bad for markets.

Markets instead rose the morning after Trump’s win, causing Bankman-Fried’s employer to have one of its largest single-day losses in its history, according to the book. Bankman-Fried had been overconfident about the market’s view of whether Trump was the better candidate.

It might feel “obvious” that smart people are better at voting, but some studies have shown that when it comes to political bias, smart people may actually better at fooling themselves and confirming their biases. Jason Brennan cites a paper by Dan Kahan, which provides some low-confidence evidence that smart people are more willing to let a negative view of their political opponents impact their views, when it comes to politics.6

There are also some real-world examples that should give us pause.

For example, California voters support green energy, and so they elected active politicians who imposed mandates, gas taxes, and subsidies, to guide the economy in that direction. In contrast, Texas voters don’t care much about “green” energy, so they elected politicians who mostly sat back and didn’t do much. The result:

The trend is not because Texas has more sunshine; it doesn’t.

California has somewhat more highly-educated voters than Texas does, but they may have tripped themselves up by thinking that they could plan their economy.

So many things are also just hard to predict. For example, nobody remembers that George W. Bush ran in 2000 as a dovish non-interventionist! He then invaded Iraq, with disastrous consequences.

Unpredictability even goes beyond a President promising one thing and doing another. You have backlash and “game theory” effects. For example, most Democrats were happy when Obama won re-election in 2012 – but Brennan points out that it set the stage for Trump’s victory in 2016. If one believes Democratic rhetoric about Trump being an “existential threat” to democracy itself, then, Brennan notes, in retrospect Democrats should wish they had voted for moderate Romney in 2012, just to preclude a Trump rise.

With all this, by far the best argument against voting, even in a swing state, is that you’re no better than the median voter at picking a candidate.

But how often have you heard someone give that reason for not voting?

Share

You are much better than the median voter

Notwithstanding the above, I strongly suspect that the readers of this blog (and “rationalists” generally) are better able to pick Presidents than the median voter.

That’s because a relatively high proportion of this blog’s readers: A) tolerate hearing views they disagree with B) at least try to put aside biases when deciding things C) have heard of prediction markets and other ways of reducing bias.

Speaking of which, Brennan invokes prediction markets as an thought experiment for why you shouldn’t vote. Here’s his logic: “If you are not able to get rich from such markets, then you have evidence that you are not a reliable predictor when it comes to politics.”

But Brennan sets much too high a bar, because it’s possible for someone to be an amazing predictor compared to the median voter, and also lose money vs the wisdom of the market consensus.7

The real test should be: Let’s give you, and the median voter, each $1,000 to make bets about politics and policy with each other (not with the market.) Who ends up with more money in the end? If it’d be you, then you should vote.

My strong bet is that the majority of readers of this blog would significantly outperform the median voter on that correct test.

To be clear: Voting is not narrowly, selfishly, rational

Voting is a “social good.” As shown above, the benefit of picking the right candidate to lead the most powerful country in the world is enormous.

But the benefit to any given voter from his/her own vote is still small.

For example, let’s take our low-end estimate that picking the best candidate is worth $182 billion to the country as a whole.

$182 billion divided by the US population of 333 million is a $546 gain per American. Not nothing.

But now combine that with the fact that, even in Pennsylvania, your odds of deciding the election are 1-in-2-million.

That means the expected value of your vote for yourself is approximately ($1,153 / 1 million), or about one 32nd of a cent.

So it’s indeed not (narrowly, selfishly) rational to vote.

Now, we’re all somewhat selfish. But unless you’re a sociopath, you care at least somewhat about broader society doing well, too.

Conclusion – vote, if you live in a swing state

The above suggests that there are three things that justify voting:

• You live in a somewhat competitive state

• You try to be unbiased and informed

• You want to improve society

If you check those boxes, and could out-bet the median voter, then I hope you’ll vote (regardless of whom you pick.)

Speaking of which, who do you pick? This is a survey just for readers who managed to read the full article!

If you liked this post, please consider sharing it, hitting the “heart” to like it (the helps the algorithm show it to more people) and subscribing!

3

I made the below refinements, that don’t make a big difference overall:

Refinement for median day change:

On a median day between 2004 and 2020, the stock market has an absolute value change of 0.49%. Let’s remove that amount from each bar, so that we see the change that could plausibly be attributed to the election result:

One other thing we could try to do is adjust for the amount of surprise which markets registered at the result of each election. Here are the probabilities each candidate had going into the election, per election betting markets:

You can see that prediction markets correctly predicted every outcome except Trump’s 2016 win, to which they gave a 20% probability.

Only one outcome was virtually certain beforehand: Bettors gave Obama a 90% chance of winning in 2008.

This gives us a sense of how much candidates were already “priced in” to the stock market. For example, since markets thought Trump had a 20% chance in 2016, and election day revealed it be 100%, the change in market values should contain 80% of the value of Trump’s Presidency (compared to a Clinton one.)

if we attribute the entire excess change in the stock market to the outcome of the election the day, we get:

I left Obama 2008 out of this graph, because the result is too large, and implausible, because the market’s update on his chances was so small (10%) that the big fall on the day following couldn’t have been due to that. If it were due to his election, then, it’d imply that his election cost the US about half of its future economic potential.)

So, here, I’ll ignore the 2008 datapoint by looking at the median impact of a Presidential election, rather than the average.

After adjusting for that, the median change was from the 2020 election, where Biden’s 20% increase in odds from election night appeared to be worth about 1.71% of the stock market, indicating that his election as a whole was worth about 6.57%.

The lowest-impact election was 2016, in which Trump’s win over Clinton was valued at 1.39% of the market.

So that suggests that the value of a vote ranges from 1.39% to 6.57% of the S&P 500.

In 2024, with the S&P worth $55 trillion, that works out to $767 billion – 3.61 trillion.

You probably should be.

That’s my new takeaway after testing OpenAI’s new model, “o1”1 this morning. It blows all other AIs out of the water on the Norway Mensa IQ test:

I had become blasé about AI progress after my initial tests in February, because there was approximately zero IQ improvement since then.

This week, that all changed.

Specifically, o1 got 25 out of 35 IQ questions correct, far above what most humans get. You can see its answers to every question here, and below some specific examples:

Examples of o1’s answers to IQ questions

Here’s the hardest question on the quiz, which it gets right:

The correct answer is probably not obvious to you. It’s certainly not obvious to me.

o1 gives this rationale:

Option D — By analyzing the patterns in the grid, we see that each column combines certain components to include all elements (V, upside-down V, rectangle sides) by the third row. In the third column, components Rleft and Rright (the rectangle's sides) have not yet appeared. Therefore, the correct answer for the empty cell is Option D, which includes the normal 'v' and the two sides of the rectangle, completing the set of components for that column.

So it says it picked “D” because it noticed a pattern in which each column should contain all possible elements. In order for the 3rd column to contain all elements, it needs the sides to be in the bottom row, which are in “D”.

It neglects to explain why it picked “D” over “B” and “F”, but a YouTube explainer I found says that’s the correct choice because "each of the 8 pictures is described only by the repetition in the diagonals ... B and E contain extra lines, so D is the correct answer."

The YouTube explainer also gives a different logic (focusing on diagonal patterns) than the pattern o1 picked up on, and I almost wrote here that o1’s description was wrong. But it seems to me like the simpler pattern it detected is also valid.

Subscribe now

Now let’s go to a question that’s more at my level:

OpenAI o1 correctly answers and explains:

In the third row, the pattern involves combining the lines from the first and second cells and removing any common lines. The first cell has lines to 4:30, 6:00, 7:30, 9:00, and 10:30. The second cell has lines to 1:30, 3:00, 6:00, 7:30, and 10:30. The common lines are at 6:00, 7:30, and 10:30. Combining the unique lines from both cells (1:30, 3:00, 4:30, 9:00) gives us Option F.

That’s exactly right!

o1 also gets some questions wrong. The following question involves the exact same pattern as Q24 above:

But this time, o1 only manages to see the additive part of the pattern, and it misses the “remove common lines” part. So it gets it wrong:

In the grid, the pattern suggests that combining the features of the first two cells in each row results in the third cell. In the third row, combining the incomplete square with vertical and horizontal lines (third row, first column) and the complete square (third row, second column) gives a complete square with both vertical and horizontal lines crossing the midpoint. This matches **Option E**.

It’s not infallible, yet! See this footnote if you want to see the correct answer.2

The IQ improvement is NOT because the questions are in the training data

The most common concern with the above kind of analysis is that maybe some AIs have a “cheat sheet”; maybe they’ve been trained on these very questions.3

To deal with that possibility, in April, a commenter on this blog, Jurij, who is a Mensa member, offered to create new IQ questions for me. I then created a survey consisting of his new questions, along with some Norway Mensa questions, and asked readers of this blog to take it. About 40 of you did.

I then deleted the survey. That way, the questions have never been posted to the public internet accessed by search engines, etc, and they should be safe from AI training data.

Using blog readers’ survey responses, I was able to align the difficulty of the new offline-only IQ quiz, and the Mensa Norway one, so that “100” should mean the same on both tests.

Then I gave the new, offline-only test to AIs. The AIs do significantly worse on this from-scratch offline-only test:

But, importantly, the magnitude of o1’s lead is roughly the same [edit: still huge]! That suggests that o1 represents a massive real improvement in AI reasoning ability, not merely the inclusion of IQ-specific training data.

Further research could help with precision of scores

Since AI vision is not yet good enough,4 it’s possible that one reason the AIs do relatively poorly on this new test is because the verbal descriptions of the questions are not written as well as they could be, and further research could include multiple wordings of the questions, written by different people, to get a sense of how sensitive the AIs are to how the questions are described.

I would also like to do further research to establish where the human average really is on these quizzes, because, when readers of this blog took the Mensa Norway questions, they averaged just 103. While I suppose it’s theoretically possible that readers of this blog who voluntarily take an IQ test are just super curious people, but not too much better at pattern analysis than average, it’s a lot lower than I’d have expected.

I asked the creator of Norway Mensa, Olav Hoel Dørum, how his test was normed, and he said it was based on giving his questions to 80 humans who had official IQ scores.

While these norming methods should get us somewhere, it would be worth norming these tests on larger populations centered around average human intelligence. That would give us a more precise comparison between humans and AIs.

For now, I think it’s a good estimate that the leading AI IQ probably now lies somewhere between 100 and 120.

Share

o1 passes a quick sanity check using “common sense” questions

In my first (outdated) post on AI intelligence, I wrote about two cases where ChatGPT-4 failed me:

I re-asked these questions of o1. It got them both right. I also tried to trick it by telling it that it was wrong, and suggesting instead the wrong answer that ChatGPT-4 had once given. It resisted my pushing and reaffirmed its correct answers.

Subscribe now

Conclusion — the future is upon us

The new o1 model shows that fast projections about AI progress are on track.

In my initial major AI IQ analysis, I extrapolated from Claude’s improvement trend to see where AI IQ might soon end up:

… look at the consistent progression:

• Claude-1 was hardly better than random. It got 6 answers right, giving it ~64 IQ.

• Claude-2 scored 6 additional points per test (worth ~18 IQ points).

• Claude-3 scored yet another 6.5 points, worth ~19 more IQ points, bring it up to above the human average.

… now consider the release dates …

• Claude-1 March 2023

• Claude-2 July 2023 (4 months production time)

• Claude-3 March 2024 (8 months production time)

A very simple extrapolation suggests that we should therefore expect to get Claude-4 in 12 - 16 months, and that it should get about 25 questions right per test, for an IQ score of 120. [OpenAI’s o1, now 6 months later, got exactly 25 out of 35 questions right]

After that, in another 16 - 32 months, Claude-5 should get about 31 questions right, for roughly 140 IQ points.

After that, in another 20 - 64 months after that, Claude-6 should get all the questions right, and be smarter than just about everyone. That’s 4 - 10 years out in total, adding up all the time periods.

Of course, that progress is not a given. Anthropic could run up against budget constraints, energy constraints, regulatory constraints, etc.

I now think that timeline is holding up, and that OpenAI has always been about 6 months ahead of Anthropic/Claude, behind the scenes.

If so, then we should start seeing AIs breaking 140 IQ in 2026.

That’s soon! Hold onto your hats.

For the record, I was quite surprised by these test results, after having been lulled into a false sense of complacency after there was no AI progress beyond IQ 100 for the 6 months since I started tracking this.

I also think all this makes clear that AIs are intelligent, and are reasoning. They do have access to all of the world’s knowledge, but they’re also more than that.

Contrary to one strand of conventional wisdom, AIs aren’t merely regurgitating words pulled out of an algorithm. Yes, they are fundamentally doing that — but predicting the next word gets so complex that logic and reasoning seem to arise out of the process of prediction. Is that maybe also the same process from which human higher-order intelligence originated from? Unclear, but there’s no doubt that seeing the dawn of AI intelligence is starting to give us some hypotheses about our own.

I’m also newly optimistic again about AI-related stocks and chip-makers, even after their recent run-ups in price. AI is also energy intensive, and so perhaps those companies could do well, too.

If you find research like this useful, and you have extra disposable income, consider subscribing below to support it! It cost me $1,000 just to get automated (API) access to o1, for example, so any offset in operational costs is appreciated:

Subscribe now

You can also help just by subscribing for free, hitting “like” on this post, and sharing it!

Share

P.S., if you’re skeptical about how smart OpenAI’s o1 is, I encourage you to try it for yourself. Paying users of ChatGPT can find it here, as o1-preview:

From following it for years, I’m well aware that there are persistent polling biases in certain states. But which states? How much? For which party?

So, I graphed it out. The yellow bars below show the current polling average in each state, where above zero means Trump leads, and below zero means Harris leads.

The blue bars show what happens if you adjust the current polling average based on polling errors (the difference between the RealClearPolitics polling average, and the actual voting results) from previous years:

My adjustment is very simple: if previous polls underestimated Trump’s margin by 4 points, then it simply adds 4 points to his 2024 estimated margin. The different blue bars all show different versions of the adjustment: dark blue adjusts for the error in 2020, light blue for 2016, and regular blue for the average error in both years Trump ran.

Past polling error hints that Trump is in slightly better shape than the raw RealClearPolitics average shows (which also shows him with razor-thin lead.)

In particular:

Main takeaway: Democrats should be more worried about Wisconsin

IF the polling bias of the last two Trump runs says anything about polling bias this cycle, then Harris is in trouble in Wisconsin, contrary to what the betting and Nate Silver’s model suggest (they both have it leaning slightly for Harris.) Note that Silver’s statistical model does not consider polling bias in the way I am here.

In both 2016 and 2020, the polling average overstated Trump’s opponent’s margin by more than 5 points. Biden still barely won Wisconsin in 2020, but the polling average has predicted Biden would win it easily.

Nevada, Arizona, and Georgia polls have not shown bias when it comes to polling Trump

We can take those polling averages pretty much at face value.

Pennsylvania and North Carolina sometimes slightly underestimate Trump

However, the bias was less in 2020 than in 2016, and in Pennsylvania the bias was effectively zero in 2020. Trump should be at most slightly encouraged by previous polling bias in these states.

Subscribe now

Historically, would it have helped to account for polling bias like this?

Implicit in my ultra-simple “model” here is that Trump is a unique phenomenon. I assume that polling for Mitt Romney in 2012 is irrelevant in the era of Trump. I also assume midterm polling bias is irrelevant. Trump brings out a unique kind of low-turnout voter, and furthermore is so controversial that it could cause some voters to not tell pollsters about their true voting intention.

But considering just Trump-specific races, we do still have one election to back-test my method on. I wanted to know: Would adjusting the 2020 polling for 2016 bias have helped predict the 2020 outcome? Would it have been better than just using the raw 2020 polling average?

Here’s a chart on that:

The graphic shows that adjusting 2020 polls for 2016 bias actually increases absolute prediction error slightly while at the same time significantly reducing bias.

In other words, the blue bars have greater total area than the yellow bars — the polling-bias-adjusted estimate was less precise at predicting the outcome. BUT, the blue bars are relatively unbiased, by which I mean they are centered around zero, unlike the yellow bars, which almost all erred in Biden’s favor.

Specifically, the unadjusted polling average underestimated Trump’s margins by 1.75 points. Adjusting the 2020 polls for 2016 bias led to predictions that only underestimated Trump’s 2020 support by 0.35 points. That means that the overall polling bias was about equally bad in 2016 and 2020.

However, the absolute error for the unadjusted polls compared to the actual election results, was 1.9 points. Adjusting the polls for 2016 acutally increased the average state’s error slightly, to 2.1 points.

We can see where that comes from by looking into the details: basically, 2020 polls in Nevada and Pennsylvania were spot on. Assuming they’d retain their 2016 bias would actually have make one’s predictions worse.

On the other hand, Wisconsin’s 2020 polls were way off, by 6 points, and adjusting for 2016 woud’ve made one’s prediction much closer.

For 2020, the best principled adjustment strategy would actually have been to not adjust polls by state, and instead just give Trump an extra 1.4 points in every state (that was the average 2016 bias.) That yields an average absolute error of 1.55 points.

While that’s what would have worked best to predict 2020, it’s a small sample size, and personally I can’t shake the sense from these graphs that something unique is wrong with Trump polling in Wisconsin.

Share

Conclusion

Previous polling bias doesn’t seem to have much predictive power, but Wisconsin is such an outlier that I wonder if bettors and modelers are underestimating Trump’s chances there.

Pollsters in Wisconsin have now badly botched two Trump elections in a row. Will that hold for a 3rd year? It’s possible that the pollsters there have learned their lesson, and are now weighting Trump demographics more heavily. In that case, the poll error adjustment could mislead.

It’s also possible that it’s now more socially acceptable to support Trump, and that the “shy voter” effect is going away. As Silver notes:

declaring one’s support for Trump has become more socially acceptable among certain subgroups, like Silicon Valley or crypto types or younger voters of color.

That could also be a reason not adjust based on previous polling error.

But overall, if I were a Democrat, past polling bias would make me worry more about Wisconsin.

The American founders designed the system to make sure each state was properly represented, and they succeeded. But an inevitable side effect is that individual votes within each state are not equal, because if most people in your state agree with you, then your opinion matters little.

Value of a vote, by state

The big, obvious factor in determining the value of a vote — more specifically, the chance of a vote determining an election — is how close the election is in a state. But another big factor is the correlations between how different states tend to vote, as well as discreet number combinations.

For example, if Pennsylvania, Wisconsin, and Michigan all tend to vote the same way, then that can dramatically increase the importance of the single electoral vote in Nebraska’s 2^nd Congressional district (Omaha), which would then control the balance of power.

Calculating that factor is not easy. I tried to do an estimate earlier this year, but didn’t get enough time. I’m a fan of Nate Silver’s approach, which is to run 40,000 election simulations, based on historical correlations and current polling data, and figure it out that way. I’m glad he’s done that work, but he hasn’t mapped it out, so I’ll do that here:

Here’s another way to visualize the value of a vote in each state (votes in states not listed are worth approximately 0):

As you can see, there are massive differences between states. Many of them you probably already knew about.

But I bet you didn’t know that a vote in Pennsylvania is worth 3 times more than a vote in Michigan or Georgia.

Or that a vote in Nebraska’s 2^nd Congressional District (Omaha) is worth more than any vote outside of Pennsylvania or Wisconsin.

Share

Using this info for campaigns, and individual decisions

These numbers could help election campaigns direct ad spending and campaign events.

They may also help us individuals make decisions. If you live in one of the green states (and especially Pennsylvania, Wisconsin, Michigan, or Nebraska CD-2) it’s probably worth voting, thinking hard about your vote, and picking the best of the two major-party candidates.

On the other hand, if you live in a state colored white on the above map, there’s little reason to care much about your Presidential vote (though there may still be important non-Presidential races.)

Analyzing candidate spending

Are the campaigns spending rationally, in line with the estimated value of a vote in each state? Roughly, yes. The WSJ made the following graph showing how much each campaign, and its allies, are spending in each state, per electoral vote (which is almost entirely proportional to individual votes):

In general, the campaign spending lines up very closely with Silver’s estimate of the value of a vote in each state.

Here’s another way to consider the same thing:

Now we can take the ratio of that amount spent per thousand votes, and Silver’s estimate of the value of each vote. I also normed the below numbers so that 1 = the average amount spent in a swing state by the party and its supporters:

This shows that, compared to Silver’s estimates, both parties are over-spending in Georgia, and under-spending in North Carolina.

It shows that, for example, Republicans are spending 72% more in Georgia, compared with their other spending, than Silver’s model would advise. It shows that Democrats are spending only a third as much in North Carolina as the model calls for.

Republicans appear to be under-spending in Michigan, and Democrats over-spending in Wisconsin.

Of course, one could also take the view that the political actors know better than the model where votes are most valuable.

It’s also possible that factors besides the value of votes are in play. For example, the media markets in Georgia could conceivably be cheaper per vote than in Pennsylvania, and that explains the difference. Relatedly, there could be diminishing marginal returns to spending, and Pennsylvania is already close to “maxed out” in terms of ad saturation.

Those are some reasons why spending could differ from the model, and still be rational.

In general, I’d say ad spending reflects the model pretty well — the rank order of spending is pretty similar.

Nebraska’s 2nd Congressional District is overlooked

Silver’s analysis now lists Nebraska’s 2nd Congressional District as having the second-most valuable votes in the country, after Pennsylvania. One extra electoral vote is particularly important this year, largely due to the following not-unlikely combination which leads to a 270 - 268 win for Harris:

In that scenario, if Trump were to win Nebraska CD-2, then it would be 269-269 — and Trump wins in a tie, because Republicans will have more state delegations in the House.

How much, in dollars, is your vote worth?

Let’s get a very rough sense of how much a vote is worth.

First, let’s consider a “revealed preference” estimate, looking at campaign spending. The WSJ reports that Republicans and Democrats have spent a combined $42 million on the Presidential election in Pennsylvania in just 3 weeks (July 22 - August 12) or $6 per voter. That’s going to be a tiny fraction of the total spent there; if past trends are any guide, the total amount would be at least 10x higher in the time until the election. So that’s $60 per voter.

Most voters are locked in, so a marginal vote is probably worth still much more than that to campaigns.

That represents the value to the campaigns. But if there’s a real difference between the candidates, then the value of the election is higher to the country, than to the campaigns themselves. For example, let’s say that the better candidate will make GDP 1/10th of 1% higher. That’s $25 billion. Over the 57 million voters in all the above swing states, that’d lead to an average value of something like ~$500/vote, just on economic issues.

Those are very ballpark numbers, but it suggests that, if one has any confidence at all that one candidate is better than the other, and you live in certain states, and you are a more informed voter than average, then your vote is quite valuable to society

A similar logic probably applies to the House and Senate — many states have competitive races this year, and overall control will be close.

Share

While I think my reasoning was valid about Biden’s mindset and incentives at the time my post was written, I missed a few things:

• All of the very top Democrats (Obama, Pelosi, Schumer, Jeffries) turned out to be far more coordinated and decisive behind the scenes than I expected them to be.

• Biden got Covid, giving top Democrats an opportunity to push while Biden was weakest.

All the post-dropout media reports I’ve seen say that Biden was privately furious at the pressure being put on him, but ultimately gave in anyway, probably due to a combination of worsening polling, plummeting fundraising, the unanimity of top Democrats, and maybe being sick with Covid.

It still seems to me that Biden could have pushed through and won, and forced top democrats to rally around him. But it seems that at his age, and in his condition, he didn’t have the energy to do that.

It also has come out that Biden campaign internal polling showed Harris with higher favorable numbers than Biden; that make have made a difference in Biden’s thinking. In Biden’s formal drop-out speech (which naturally was more about image than reality) he didn’t directly mention either his declining abilities or his age, but rather the need to “unite my party” to “save democracy” and said that the best way to do that was to “pass the torch to a new generation.”

In the end, the betting market consensus had a better understanding of the abilities and determination of top Democrats than I did, so let this be a lesson to listen to the markets.

Reminder that markets in general have a good track record

On ElectionBettingOdds.com, I’ve been tracking elections for nearly 10 years. Across hundreds of different races (for President, Senate, House, primaries, state-by-state, international, etc) we can get a picture of bettors’ accuracy.

In the graph below, the white line shows what predictions would look like if they were perfect. The blue line shows how often bettors' predictions within different ranges (e.g., between 60-80%) actually came to pass. Where the blue line differs from the white line, the predictions were off a bit:

At the time of my post, markets gave Biden a ~52% chance of dropping out.

As you can see, historically, when markets said something would happen within the 50-60% range, it actually came to pass about 57% of the time.

Subscribe now

So I should have had more humility when predicting that the real chance was 15%. While all of my reasoning may have been valid enough, as mentioned above, it missed some possibilities on the other side, such as just how determined actors like Obama, Pelosi, and behind-the-scenes congressional Democrats were to push Biden out.

By the time the news had broken that all of those Democrats were pushing so hard against Biden, the betting moved to give Biden a ~74% chance of dropping out, and I adjusted my view:

I now see little reason to disagree with the bettors, whereas last weekend, I thought Biden would stay, with high confidence. My change in thought is because it has been revealed that just about every top-level Democrat is pushing hard. I mistakenly underestimated how coordinated the Democratic Party was. Biden getting Covid also didn’t help things. I still have outstanding bets on Biden staying, and I haven’t decided if I’ll wrap them up, because the odds seem pretty plausible right now.

I did not end up closing out my bets on Biden, since the odds still seemed not too low. I’m still “in the black” in my long-term betting record, but this was the biggest miss.

What are the markets saying now?

Now that we’ve reminded ourselves to listen to betting markets rather any single person (even with money on the line) let’s check in on what the markets are saying right now:

1- Harris’s electability is on the higher end of what bettors previously expected

ElectionBettingOdds.com has a page showing the chance of winning the Presidency IF a candidate is nominated. Below is the graph for that, zoomed in on Harris’s blue line:

In general, the betting markets thought she’d have a 38% - 48% chance of winning, if nominated. Currently, bettors think she has a 48% chance of winning if nominated.

I think the biggest reason for that shift is that Democrats coalesced very quickly. Harris didn’t have to re-litigate her often-extreme positions that led her to being labelled the most left-wing senator in 2019. She didn’t have to field attacks from other Democrats or the media. Instead, it was a perfectly smooth path, with virtually no criticism from either within the party, or from legacy media outlets.

Which, again, is a testament to the coordination abilities of top democrats. They may not be known for picking candidates in the most democratic way, as candidates for both 2016 and 2024 were essentially picked by party power-brokers, but they do seem coordinated and effective.

2- Betting markets say Harris’s VP will probably be Pennsylvania governor Josh Shapiro

I don’t know much about Shapiro, but he does have a high approval rating in Pennsylvania (somewhere between 54% and 61%) and is considered competent and not radical, which could help offset Harris’s negatives.

There are some concerns that he’ll offend some Democrats due to his strident support of Israel.

I think he’d be Harris’s best pick from an electoral standpoint, because of the state map:

Subscribe now

3- The “blue wall” is up for grabs

The three southern/western “swing states” (NV/AZ/GA) definitely have a lean, per the bettors, with Trump favored with a 60%+ chance in each.

However, the old “blue wall” states of WI/MI/PA are all up for grabs, and are coin-tosses. Trump needs to win all of the southern/western swing states plus ONE of those to win, which is currently the modal (single most likely) outcome:

However, because Pennsylvania has the most votes of those three, if Trump wins just Pennsylvania, he can then win with JUST Georgia, or JUST Nevada and Arizona. I listed the various combinations in more detail here.

That’s why Nate Silver estimates (link; paywalled) that Pennsylvania has a 34% chance of tipping the election. The next most likely state to tip things is Wisconsin at 14%.

4- Limited impact on Congressional outcomes

I find it surprising that Congressional control odds have moved little, despite the big shift in Presidential odds. The House:

The Senate:

5- Who can put their foot in their mouth less?

Trump and Harris are known for being gaffe-prone, and alienating to many.

I think the debates will be fascinating, and that the winner will be whomever can look like a calm semi-Presidential adult, and avoid musing about whether Harris is “really black” or whether she “slept her way to the top.” And to avoid making shrill allegations about how Trump will “end democracy” and how Trump called Nazis “very fine people” (he didn’t.)

Trump can win if focuses on Harris’s track record overseeing unprecedented chaos at the border, high inflation, and her views like wanting to ban fracking in 2020, among many others. And Harris has plenty of real Trump flaws to hit him on.

It’s going to be close!

The last week has brought rapid swings in the betting odds. Here are some of them:

Biden will (probably) drop out. The odds have shifted massively

Here are the current nomination odds, including the graph for the last month:

An amusement park should base their next ride on that graph!

The latest plunge in Biden’s odds reflects media reporting that virtually the entire elite of the Democratic Party has united to force Biden out of the nomination. The list of officials pressuring him out now include Barack Obama, former House Speaker Nancy Pelosi, Senate Majority Leader Charles Schumer, and House Minority Leader Hakeem Jeffries — with Schumer and Jeffries speaking not just for themselves, but for Democrats in the Senate and House in general.

Leaks and anonymous sources suggest that Biden may finally be bending under all that pressure, and may be preparing a dropout speech for as soon as tomorrow or Monday. Additionally, Biden has Covid right now, which might have taken a bit of fight out of him.

In an interview with BET earlier this week — after congressional leaders asked Biden to drop out — Biden had already shifted his tone slightly, saying he was reluctant to drop out. That was a bit of a change in tone from his earlier strident insistence.

I’ve also noticed that Biden’s spokespeople are saying things like “he IS running” rather than “he won’t drop out.” A very slight difference, but I think they’re using “weasel words” to avoid overtly lying.

I now see little reason to disagree with the bettors, whereas last weekend, I thought Biden would stay, with high confidence. My change in thought is because it has been revealed that just about every top-level Democrat is pushing hard. I mistakenly underestimated how coordinated the Democratic Party was. Biden getting Covid also didn’t help things. I still have outstanding bets on Biden staying, and I haven’t decided if I’ll wrap them up, because the odds seem pretty plausible right now.

The case for Biden staying rests entirely on the fact that he could still probably get the nomination if he wants. But does he want to fight his entire party on the matter?

Maybe. He just issued a statement again saying he’s staying in the race.

We’ll find out.

The betting shifts over the last month have been quite spectacular. Roughly:

— Debate night, June 28: Biden plunged from 85% to 60% — most of it while the debate was still happening,

— Week after the debate: The market mostly retained its post-debate level.

— July 2: There was another huge plunge, from 65% to 35%, as the first elected Democrats started to turn on Biden.

— July 8: Biden’s odds recovered as he launched a concerted media push, and firmly insisted he wouldn’t quit

— July 10: Biden’s odds fell to the 30s again, on news that Congressional Democrats were plotting to coordinate an effort to get Biden out

— July 12 - 17: Biden’s odds began a gradual rebound, after he gave a relatively strong NATO press conference, without a teleprompter, and as he started to “run out the clock” to the convention. His odds rebounded all the way into the 70s.

— July 18: Biden’s odds collapsed as he tests positive for Covid and, simultaneously, news broke that the previously-mentioned top leaders all met with Biden and told him to drop out. Biden’s drop further continued as reports cited anonymous sources claiming that Biden was finally giving in, and would likely announce dropping out either Saturday or Monday.

Biden’s odds hit a low of 11% yesterday, before returning to about 30% today, as Biden issued a formal statement saying he was still running, and a few others backed him up, including AOC (which I wouldn’t have predicted). There’s a chance the anonymous reports are wrong, and at the end of the day, it’s up to Biden whether he’ll be the nominee.

Thanks for reading Election Betting Updates! Subscribe for free to receive new posts and support my work.

Trump leads, but not by much

Bettors slightly favor Trump in the election:

Trump saw a debate boost taking him from about 55% to 60%. Then he saw an assassination attempt boost which brought him to nearly 70%.

But then that gain evaporated as it looked likely that Biden would drop out, and that Democrats would field a stronger candidate against him.

Betting markets are clear: Democrats would be slightly better off with Harris, and much better off with Newsom

Looking carefully at the above chart, one can see that Trump’s recent decline coincides pretty precisely with Harris’s recent rise, and Biden’s fall.

That’s what we should expect from the conditional odds, which have consistently said Harris is more electable (though the margin is unusually tiny at this moment):

Some of Trump’s decline might be because of rumors of an open convention, which makes bettors particularly optimistic about Democrats’ chances.

Economist Adam Ozimek also did a regression analysis on historical betting data, and found that Biden dropping out would add about 6 percentage points to Democratic chances of winning the Presidency:

Congressional odds move with the Presidential odds

For the most part, the odds on House control cling stubbornly to 50/50, but they do shift a bit as a party’s Presidential chances shift:

The Senate odds shift even less:

So the bettors say that Democrats should not expect the top of the ticket to matter too much for control of Congress. Local factors, ad spending, and candidates, will be more important.

Share

Conclusion

It’s a wild election!

I’m very curious what will happen in the Democratic primary, which is a historically unprecedented situation. ElectionBettingOdds.com is the best place to get a single summary of number of what’s going to happen.

Now, I’m going to post this before I have to redo all the screenshots again.

As the creator of ElectionBettingOdds.com, I almost always defer to the betting markets. They’re the single best predictor.

But just as with the stock market, they aren’t perfect. People do make money by having insights that markets miss.

I think the markets are wrong when they say it’s a coin toss that Biden will drop out:

My personal view is that the odds should be 85%+ that Biden will be the nominee.

I have no respect for pundits blathering without putting money on the line, so rest assured that I’ve bet as much as I can conscionably bet on this. It’s also my only bet so far in the 2024 cycle, because I generally think the markets are right.

Here’s why I’m confident Biden will stay:

Biden’s incentives: the race is very winnable for him

So much analysis about whether Biden will drop out focuses on:

• A) whether it’d be good for Democrats and

• B) whether prominent Democrats want Biden to drop out

I’ve seen very little analysis about whether it’d be good for Biden for Biden to drop out. But that’s by far the most important question, because it’s Biden’s choice whether he drop out — Democrats can’t force him to.1

So, what would be good for Biden? What are his chances of becoming President? Here are the “conditional odds” from my site, which I agree with:

As you can see, Biden is indeed the least-electable candidate.

So Democrats, and their supporters in the media, are not crazy to lead a charge to switch horses.

It’s also worth noting that the ordering in those odds has been quite stable since the debate, as you can see from the chart. Bettors are confident that Newsom would be more electable than Harris, who also tends to be ~10 points more electable than Biden.

If you’re a voter in a Democratic primary, that is an excellent reason to vote against Biden as your nominee.

But now imagine that you’re Joe Biden, and you already won the delegates required for your party’s nomination.

Now a bunch of pundits and swing-seat Democrats are asking you to step down, which almost certainly would mean Harris gets the nomination, which would make your party a mere ~10 percentage points more likely to win.

So Biden is supposed to trade a 34 percent chance of being the next President, to give his party a mere 10 percentage point gain in their chances against Trump?

Come on, man.

Politicians are self-interested.

Biden did not really run to “stop Trump”, contrary to his talking points. That’s a fig leaf. You can tell that from the fact that he ran not just in 2020, but also in 1988 and 2008. Biden has always been an ambitious guy whose life dream is being President. He wouldn’t have the job if it weren’t.

The above is the rational, market analysis. But of course, Biden probably isn’t checking ElectionBettingOdds.com (though I’ve heard that Vivek Ramaswamy and Marco Rubio have visited it.)

Biden instead looks at polls:

Polling shows an even weaker case for Biden dropping out

Here’s the current RCP polling, which consists entirely of polls taken after the debate:

Yes, Biden is down, but 2.9 points is hardly a historic disadvantage.2

Before the debate, Trump led by 1.5 points:

So that’s a mere 1.4 point shift. That’s small!3

How much better does Harris perform? Hardly at all:

So Biden is supposed to look at those numbers and say “well my Vice President now does 0.7% better in the polls against Trump, guess I’ll pass the mantle.”

It’s ridicule-worthy from his perspective, honestly.

Admittedly, Biden does worse in the electoral college than in the popular vote, so the picture is a bit bleaker for him than the above polling suggests. The average for Pennsylvania, the most likely pivotal state, currently has Trump up 5, which is substantial. But it’s not obvious Harris would do better. The latest poll that asked about her in Pennsylvania was in February, by The Hill/Emerson, and it found Trump beating her by 9, and beating Biden by 2.

Also, considering things from Biden’s perspective, the same pundits who are freaking out now are the same ones who underestimated his chances in both 2020, and 2022, and probably in 2016 (there’s no way to know sure for if Biden would have won if he’d ignored advice from partly elders and run against Clinton and then Trump — but he likely would have.) He clearly has a disdain for his own party’s pundits and politicos, and I don’t blame him.

The calls to drop out are mostly about surprise and panic among Democratic insiders

As both the polls and betting above show, the fear among Democratic politicians and journalists is far less reflective of the fundamentals of the race, than it is about their own mindsets.

Specifically, many of them believed4 the partisan spin in the New York Times and elsewhere that Biden was “sharp as tack” and which condemned (real) viral videos of Biden gaffes as “cheap fakes.”

The debate disabused them of that spin, and suddenly they realized they were running an increasingly-senile candidate in what many think is the most critical election to preserve our democracy (I don’t think that, for the record.)

Perhaps partly because of the sudden nature of their realization, many Democrats have now overcorrected and are too negative about Biden’s ability to win.

Share

Biden is getting senile, but is more cogent than current media coverage would make one think

Biden has had declining verbal abilities throughout his entire term, and it’s been a slow decline. Based on that, I predict: he’s not going to be a “vegetable” by November, as some fear.

Furthermore, while Republicans might not like to hear this part, I think Biden’s logical thinking ability, and sense of himself, is still functional.

His verbal ability has long been on the fritz, and he misuses words, slurs his speech, and constantly loses his train of thought. But I haven’t yet seen an example of his logic being impaired.

For example, the other day, Biden said “Vice President Trump” instead of “Vice President Harris.” Embarrassing! But it was also clear from the context of his sentence that he didn’t actually momentarily think Trump was his Vice President. He didn’t start wondering if maybe his orange VP could mortgage some properties to top up the campaign war chest. Instead, it was just a disconnect between his brain and his mouth.

That’s not an endorsement of Biden, and verbal reliability is obviously useful for a President, but I think it is worth trying to separate those two aspects of cognition, when listening to Biden.

Additionally, I see no indication that he’s losing the drive, or his mentality that got him into the most powerful job in the world.

What about his legacy? He’ll go down in history as the candidate who lost to Trump!

The main point I’ve seen pundits make that touches on Biden’s own interests goes like this:

• If he drops out now, he’ll be seen by Democrats as a George Washington -like figure, putting his ego aside to retire to his metaphorical farm, and saving the country in the process.

• If he persists, he’ll be seen as the doddering old fool whose vanity ushered in an orange tyrant, destroying American democracy.

But there are a lot of assumptions built into that framing, and while they are shared among the Democrats’ core base, I see little indication that Biden shares them.

First, in terms of legacy — does it actually look so great to drop out after one bad debate? That’s kind of embarrassing, too. And what if Kamala Harris loses to Trump, which is also likely? Then Biden will not only be a quitter who admitted senility and dropped out in the middle of a campaign for the first time in US history, but he and others would probably wonder if he could have done better than Harris in the election, and whether he failed the country that way.

People at Biden’s level also just generally don’t quit while they have a chance. If people at that level had personalities like the rest of us, they wouldn’t be President. You can see that in the President’s mindset. From his NATO presser this week (no teleprompter):

Well, look, I’m not in this for my legacy.  I’m in this to complete the job I started.

Also from the NATO presser, a glimpse of his mindset:

I always have an — an in- — inclination, whether I was playing sports or doing politics, just to keep going, not stop. 

A lot of the pundits also show their neuroticism when they fear “oh, you’ll be seen as the person who made Trump win!” They forget the running for office, against Trump or not, is inherently risky. Politicians by definition have to be okay with a lot of personal career risk. Such people don’t frame things as “is there a chance I’ll lose?” Instead, their mindset is, “is there a chance I’ll win?”

Biden made that very clear in his NATO presser:

Q:   If your team came back and showed you data that [Harris] would fare better against former President Donald Trump, would you reconsider your decision to stay in the race?

THE PRESIDENT:  No, unless they came back and said, “There’s no way you can win.”  … No one is saying that.  No poll says that. 

Share

Steelmanning the “drop out” camp

Let’s consider a couple of points raised by those who think Biden will drop out (for example, Nate Silver is a smart guy who has predicted that Biden will drop out.)

One point Silver raises that hasn’t already been addressed above:

I don’t think there are actually that many Democrats who have an incentive to keep this going … if you read the smoke-signals carefully, so are some of the elder statesmen in the party, like James Clyburn and Barack Obama. No, they’re not saying this explicitly — but that Obama failed to refute a Washington Post story that suggested he has private doubts about Biden is what we poker players call a “tell”.

I’ll add that Pelosi also had a similar “tell” recently when she said:

“It’s up to the president to decide if he is going to run … We’re all encouraging him to to make that decision. Because time is running short.”

When she said that, Biden had already loudly, forcefully, scornfully, repeatedly, said that was definitely running and not leaving the race. So Pelosi’s “I hope he’ll decide soon” was a veiled “you didn’t really mean to choose that, did you? Decide again please.”

I just don’t think it matters.

Silver has a post titled, “Biden has a weak hand.” I see the opposite. I think Biden has all the cards, because he already won the delegates, and because the Democratic establishment (donors, politicians, etc) have literally nobody else to support if he insists on staying in (who would they support? RFK Jr.? Jill Stein? Cornell West? Nope.)

Additionally, Biden’s view is that Democrats calling for him to resign are largely playing self-interested politics. From Biden’s NATO presser:

I served in the Senate a long time.  I understand the impetus of candidates running for local office and whether they think the top — if they can help them or not. 

In my state of Delaware, which was a very — at least a purp- — it wa- — it was a red state when I started in terms of where you now talk “red and blue.” [My fact-check: 52 years ago in 1972, the year Biden first won his Senate seat in Delaware, Republican Richard Nixon won the state by 20 points. In general it seems to have been a swing state at the time.]

I — I don’t recall most of the Democratic presidents winning my state when I was a candidate.  The truth of the matter is I understand the self-interest of a candidate.  If they think that, you know, running with Biden at the top of the ticket is going to hurt them, then they’re going to run away.  I get it. 

So, Biden is inclined to discount the congressional calls for him to leave.

Another Silver argument is that Democrats are polling just fine in Senate races, and therefore, if Democrats can just run a more generic Democratic for President, they can win.

I think that somewhat overstates the value of a generic Democrat running because local candidates have local advantages5 but it’s also true that Democrats would likely win with a generic candidate — the conditional odds discussed at the start of the post suggest California governor Gavin Newsom has ~60% chance of beating Trump, compared to Biden’s ~34% chance.

Even if Biden knew that and believed that, I doubt he’d drop out and endorse Newsom.

Furthermore, getting him to believe it is nearly impossible, because current polls show Newsom doing even worse than Biden:

I think Nate Silver, and the bettors, are right that IF Newsom (or similar) were the nominee, his silver tongue and comparatively youthful demeanor would win him more support, and he’d win in November.

But because most Americans barely know who Newsom is, he lags in the polls. And good luck making that counterfactual case to Biden. “Yes, you out-poll Newsom, Joe, but he’s such a good campaigner that if you give him the chance, he’d actually beat Trump, where you’d fail!”

If I’m wrong …

I think there’s ~15% chance that I’m wrong and Biden does drop out. If he does, I think the reason will be:

1. There’s an incident (fainting at a podium, a McConnell-like 30 second brain freeze, or something similar) that is just so bad that everyone he respects, from Barack Obama to Charles Schumer, calls him and begs him to drop out. It’ll have to be pretty bad; remember that he survived having a bleeding eye in the 2020 debates.

2. Biden’s campaign team sees he’s actually much worse than the rest of us can see, and a huge number quit.

I used to think maybe Biden might himself decide that the trouble of campaigning just isn’t worth it at his age, but after watching him more, I see no signs at all of that kind of mentality, even though the typical person would feel that way at age 82.

Share

Biden in his own words

I’ll end with a couple lines from Biden’s NATO presser, in which he gave a clear window into his thinking and mindset.

And:

I think I’m the most qualified person to run for president.  I beat him once, and I will beat him again…

there were at least five presidents running or incumbent presidents who had lower numbers than I have now... 

And:

… my schedule has been full bore. 

I’ve done … roughly 20 major events, some of them with thousands of people showing up.

And so, I just think it’s better.  I always have an — an in- — inclination, whether I was playing sports or doing politics, just to keep going, not stop. 

Conclusion

I think political pundits and reporters (and bettors, for once) are totally out of touch with Biden’s thinking on this, and Biden’s thinking is what matters most.

I suspect that many are led astray by motivated reasoning — Democrats WANT Biden to drop out, because other Democrats do have slightly better prospects.

I also think there’s a failure here to try to really get into Biden’s head. Instead people assume he has similar priorities to themselves (typically, stopping Trump.) But actually actually Biden is just very ambitious, not a quitter at all, and proud of his record.

I could be wrong! It’s very possible that you should just listen to the markets, and ignore me. But I feel sure enough about this take that I decided to write it up.

On Tuesday, the site I created, ElectionBettingOdds.com, looked like this:

The odds reflected a close race. Trump was favored at 54% — but that’s still pretty close to a coin toss.

The narrow race was reflected in the map, which showed that the most likely single outcome was still a Biden win, 270-268. That didn’t mean Biden was favored overall, because Trump has a greater number of electoral paths to a win, and the swing states that leaned Trump were more safely in his camp than the ones that leaned Biden.

Bettors react slightly to Nate Silver’s model

Nate Silver released his new election model just after noon on Wednesday.

Silver is the election modeler with by far the best historical track record — it’s good enough that he (barely) beat prediction markets, as measured by ElectionBettingOdds.com, over the period of 2016 - 2022. Interestingly, taking the average of Silver’s model and ElectionBettingOdds.com was even more accurate than looking at either one.

Silver came out swinging with his new model, giving Trump a 65.7% chance:

That’s much higher than the 53.9% that the betting odds gave Trump.

I’ve often been asked — can we really know who’s more accurate, bettors or Nate Silver? After all, bettors have the option of copying Silver.

So I was curious to see how the bettors would react to Silver’s just-released model. During the course of the day, in which there was no big polling shift, bettors did update their predictions a bit:

Specifically, in the seven hours after the model’s release, bettors erased about 15% of the gap between their previous estimate, and Silver’s.

So it seems bettors do pay attention to Silver’s model, but not that much.

Debate night moved the odds more

Just 27 minutes into the Trump/Biden debate on Thursday night, bettors adjusted sharply towards Trump:

By the end of the debate, bettors had split up Biden’s lost odds relatively evenly between A) Trump, and B) Democrats who would get a shot at the Presidency if Biden backed out of the race:

After the debate, Biden’s odds fell still further, as even left-wing commentators on CNN and elsewhere panned Biden’s performance, and suggested that he should withdraw. Later, even the NY Times editorial board called for Biden to step aside.

Here’s the change in the odds, with the arrows showing the change over the whole of last week:

Interestingly, the majority of Biden’s lost chances now go to other Democrats.

Here’s what the nomination odds now look like:

Subscribe now

Conditional odds say that Democrats SHOULD try to replace Joe Biden

We can use the above statistics to estimate “conditional” odds — in other words, the chance that any given candidate will win *IF* nominated. Here’s what that shows:

Specifically, if Newsom were to be nominated, he’d probably beat Trump.

But if Biden stays in, Trump will have a 2/3rds chance. That’s no longer a coin toss.

These odds are volatile, as they combine and magnify any errors/inefficiencies in both the nomination and President markets. But over time, they do give a sense of electability.

Conclusion

This was an eventful week, and the betting odds tell us a couple of things:

— Democrats absolutely could still win.

— Even Biden still has a path to winning, but Democrats would have much better chances if they replace him

— Bettors don’t react that much to Nate Silver’s odds; historically, it’s best to take an average of his model, and the bettors’ predictions.

Share

More …

Republicans are likely to take the Senate:

House control is a toss-up:

Trump is likely to pick former North Dakota governor Burgum, who has good business experience — or Yale-educated populist JD Vance (who’s also the author of a good book, Hillbilly Elegy)

You can go to ElectionBettingOdds.com for live odds!

As I described in a recent post, bureaucrats at the CFTC (a federal agency) have officially proposed banning trading on elections, Academy Awards, Supreme Court decisions, and pretty much anything interesting.

Reacting to that post, Richard Hanania commented:

After talking with multiple prediction market founders this weekend, I’m pleased to report that legal theorists have already been working hard on this issue for several years, and we’ll soon learn if their approaches pay off. Here are their approaches, in more detail:

A federal judge just heard arguments in a lawsuit challenging the CFTC’s authority to ban election trading

The CFTC order justifies its proposed ban by calling election trading “gaming,” which is a questionable assertion. The bureaucrats call it gaming because the actual law they are supposed to enforce says that trading may not involve:

(I) activity that is unlawful under any Federal or State law;

(II) terrorism;

(III) assassination;

(IV) war;

(V) gaming; or

(VI) other similar activity

The CFTC says trading on elections amounts to “gaming” because it’s gambling. That’s off-base for two reasons:

1. The law establishes categories of trading that are forbidden, banning trading about terrorism, assassination, gaming, etc. The CFTC is wrongly interpreting the above law as forbidding gaming itself, and then declaring election trading to be gaming.

2. These markets don’t work like casino gaming or bookie odds. Rather, contracts are traded between buyers and sellers in a transparent way, just like trading works on stock exchanges. Yes, I created “ElectionBettingOdds.com”, because that’s more fun than, say, “ElectionTradingOdds.com” — but it’s betting in the same way that short-term trading on Telsa or Bitcoin or NVIDIA is gambling. In fact, many do gamble on the value of such companies, but we also recognize that it’s different from playing at a casino, because it’s formalized, fair, and provides valuable information.

Kalshi filed a lawsuit against the CFTC making those points, among others. Hearings were held just a week ago.

As Kalshi notes:

The Order’s analysis fails at every step. It contorts and misapplies the CEA’s text, ignores its structure and purpose, allows a narrow exception to swallow the rule, and engages in faulty and unsupported reasoning.

Trading an event contract can never amount to terrorism, assassination, or warfare. Accordingly, the only way to make sense of the provision as a whole is to read the CEA’s reference to contracts that “involve” those activities as focused on the contract’s underlying event. For example, the Commission may prohibit a contract contingent on whether the President will be assassinated.

Their lawsuit was filed against a 2023 decision by the CFTC to reject a Kalshi application to list election contracts — but the logic equally applies to the new CFTC proposal to ban all event contracts.

I’m biased in my sympathy to prediction markets, and not a lawyer. But to my eye, the lawsuit sure seems to make strong arguments.

Kalshi’s co-founder still has hope that the court might rule in its favor before the election, paving the way for them to list election contracts before the election. A ruling is expected between June and September.

If the judge rules in favor of Kalshi, that could force the CFTC to allow not only Kalshi election markets, but it would also jeopardize the logic outlined in the CFTC’s proposed ban, as it is fundamentally the same.

I talked with two people who have been very closely following the case, and they both put the odds of Kalshi winning at about 55% — up from what they had thought before the hearing, as it apparently went well. Play-money markets (which are admittedly very thin, with just 23 traders in total) put the odds a bit lower:

Subscribe now

Can the CFTC be swayed? Worth a try, says PredictIt

The courts are the primary hope in blocking the CFTC’s prediction market ban. If that fails, the other possibilities are A) Trump winning the election, or B) the CFTC commissioners changing their minds.

I think the odds of the CFTC commissioners changing their minds are very low, but PredictIt has emailed their traders encouraging everyone to submit public comments with the CFTC. They write:

We need you to act now to save elections markets. The CFTC is accepting public comments on its proposed rule, but only for a limited period. Time is short. Please take this opportunity to tell the commission about your own experience as a seasoned political trader and why you think they are making a mistake.

Read the proposed rule and leave your comment to the CFTC here:

SUBMIT A COMMENT

Believe it or not, the CFTC is required to read and respond substantively to every public comment. Proposed rules are often modified and, in some cases, withdrawn based on information received during the comments period. The more comments, the better!

Anyone can submit a comment. It doesn’t need to be fancy — it can just cite legal arguments like the above, or like the points Kalshi raises in their filing.

There’s one month left to submit public comments.

Share

PredictIt could win a separate legal battle, which would allow it alone to operate

Back in 2014, when the CFTC was more reasonable, it granted PredictIt a “no action” letter, which effectively allowed PredictIt to operate. Then in 2022, the CFTC revoked the “no action” letter and ordered PredictIt to shut down.

Since then, PredictIt has been suing the CFTC in court, and so far, courts have forced the government to let PredictIt stay open while a ruling is pending.

PredictIt recently won a fight over jurisdiction, so that their case will now be held in Texas, in the 5th Circuit — which tends to be friendlier towards markets.

A ruling on that case isn’t expected until next year. If that ruling goes well, then things will continue as they have since 2014, where PredictIt has a grandfathered monopoly on prediction markets. Well, not entirely a monopoly. Specifically:

• The Iowa Electronic Markets is also exempt, as it holds the only other “no action” letter from the government — but you literally need to mail a check to fund an account there, and traders are limited to $500. It’s an academic’s hobby, not a business.

• Manifold Markets has been a play-money market up till now, but in about a month, it will add a bit of a real-money aspect to the site, in which traders will win real-money prizes which are proportional to investment performance. Their strategy leans on an exemption in the law for sweepstakes, which are legal.

But aside from those things, PredictIt would again be the only major, clearly-legal operation — if Kalshi loses its lawsuit against the government, and PredictIt wins its suit. That’s not ideal, but I’d rather be able to trade at PredictIt than nowhere!

Subscribe now

Conclusion: I’m optimistic

Being at the Manifold conference this weekend showed me just how actively many smart entrepreneurs are in the space, working to counter the CFTC’s bans in court using many different strategies.

There are at least 3 separate legal approaches being tried simultaneously. To recap:

— Kalshi’s lawsuit, challenging the heart of the CFTC’s authority to ban prediction markets.

— PredictIt’s lawsuit, fighting to preserve its grandfathered status as a legal US exchange.

— Manifold’s attempt to merge sweepstakes with play-money markets.

This gives a lot of potential avenues for at least some prediction markets surviving in the US.

Considering the Kalshi case, let’s say there’s a 50% chance the courts eventually reverse the CFTC, a 40% chance Trump wins and his appointee overturns the ban, and a 5% chance the CFTC just changes its mind.

Then, the odds it will be banned are: (.5 * .6 * 95/100) = 28.7% chance prediction markets in general will be banned. That’s very close to my previous estimate of 30% — but this estimate is more informed by people close to the issue.

And within that, let’s say there’s another 50% chance that PredictIt is allowed to operate.

And also another 50% chance that Manifold becomes a meaningful real-money market using sweepstakes.

That would suggest something like a (.5 * .5 * .287) = 7% chance that prediction markets will be totally banned in the US.

That’s comforting.

In general, attending Manifest this weekend, I got the sense that there is a lot of entrepreneurial energy going into prediction markets.

This weekend, Scott Alexander also asked Nate Silver about the likelihood that, in 20 years, prediction market odds will be as central to political discourse as polls are today. Silver estimated a 20% chance that they would be as central as polls, and also an 80% chance that they would be more-cited than today.

That seems right.

But the lessons from Covid are still useful. One is that transmission is virtually non-existent outside. Could that insight help us reduce disease prevalence in general?

I’d think so. The difference between outdoors and indoors mainly comes down to airflow and UV sunlight. So with enough fans blowing new air in, or with UV light, could indoor spaces become as safe as outdoor spaces?

I first heard of the notion from Alex Tabarrok at Marginal Revolution. He wrote:

A recent paper … shows that air filtration and UV disinfection can greatly reduce SARS-COV-II in hospital wards. The authors installed portable air filters with UV disinfection on two COVID hospital wards in the UK … the portable filters and UV light … greatly reduced multiple viral, bacterial, and fungal pathogens.

Here’s the most relevant graph from the paper. The left panel shows pathogens in the air without filtration, and the right panel, with it:

If it’s so effective, why isn’t such filtration more common?

I figured cost was likely a big problem. But, digging into the study, the filter they tested in the ICU ward is not that expensive:

Certainly, it’s not expensive in a hospital system that sometimes charges patients that much for things like a single “routine nasal swab test” and for blood tests.

Subscribe now

No good randomized, experimental studies have estimated the effect of air filtration on actually catching colds

The above study shows filtration and ventilation reduce disease particles in the air. That seems a good reason to think transmission would also be lower. But there aren’t good experimental studies actually testing that. A summary of studies on ventilation notes that:

ventilation [and] air cleaning devices … may have a role in reducing transmission ... However, the evidence was usually of low or very low quality …

One of the less-weak studies surveyed over 100 schools, and found:

Adjusting for county-level incidence, COVID-19 incidence was 37% lower in schools that required teachers and staff members to use masks, and 39% lower in schools that improved ventilation, compared with schools that did not use these prevention strategies.

Ventilation strategies associated with lower school incidence included … opening windows, opening doors, or using fans (35% lower incidence), or in combination with … (HEPA) filtration … 48% lower incidence.

The study wasn’t experimental/randomized, so there are likely confounding variables (for example, the schools that took these measures may have also taken un-measured precautions.) But basically, the best available (weak) evidence suggests that just opening windows is as good as making everyone mask up, and that adding an air filtration unit is better than that.

Share

It is clear that Covid and the flu don’t spread much outside

A study in the Journal of Infectious Diseases from 2021 notes that:

the odds of indoor transmission was very high compared to outdoors (18.7 times; 95% confidence interval, 6.0–57.9).

So outdoor transmission is between 6 and 58 times less than indoors.

That seems plausible. I came around to the ultra-low risk of outdoor spread when the Black Lives Matter protests and riots first began, and health officials reversed their guidance against public gatherings (as long as the gatherings were about anti-racism.) While an insane policy, I was eager to see what would happen to Covid cases — would they spike, with everyone out breathing on each other in the streets? Over the following months, the data came in: Covid cases did NOT spike.

Studies on the flu are scarcer than for Covid, but there is evidence of reduced harm outdoors. The study mentioned above summarizes another:

Summers et al. conducted a historical analysis of a large outbreak of the 1918 influenza virus on a military troop ship in July 1918. The outbreak involved over 1000 of the 1,217 crew members and caused 68 deaths. Analysis of factors that might have contributed to mortality revealed a significant association between individuals who slept indoors, in cabins with bunks (mortality of 146.1/1,000 population), versus individuals who slept in hammocks in open-air areas (mortality of 34.1/1,000 population). This study is of particular interest because the duration of exposure and distance between individuals was held constant …

Subscribe now

Cost of the common cold

This is an important question, as anyone who’s had a cold can relate to. But let’s quantify it anyway: The latest study I could find on the cost of the common cold put it at $40 billion in 2000. Adjusting for GDP growth, that would now be $107.83 billion.1

That’s just under half of 1% of total economic output. The effect on happiness is probably even more important, as the bottom of this graph cited in last week’s post suggests:

Why aren’t places doing more ventilation?

I think most people are just unaware of how helpful ventilation appears to be for preventing cold transmission.

Even if people are aware of the benefits, there may be a bit of a collective action problem. But step one is people actually valuing ventilation properly.

Before businesses, hospitals, etc, will pay money to set up good ventilation, their executives need to know that it works. It also helps if customer know, and can prefer to go places with ventilation.

People should also do more studies to confirm this, but it sure looks like it’s very effective, from what we know.

Conclusion: Covid taught us that being outdoors, or ventilating indoors, was great for preventing diseases. Let’s learn that lesson!

While society is waiting to catch up, there are things people can do as individuals:

• Is someone coughing behind you on the airplane? Try turning on the overhead air vent to blow some of it away. Planes actually have excellent air filtration systems.

• Seeing someone with a cold? Meet outside.

• Is your Uber driver sneezing? Make sure a window is cracked open.

… etc.

Of course, fresh air is nice anyway, unlike masking — so this solution is really a win-win from all perspectives.

P.S.: regarding colds: Studies find that zinc lozenges reduce cold length and severity significantly, by a third. [Update: Commenter Milli notes that there is a small randomized trial showing no effect. Then again, a LessWrong poster notes that the trial used zinc lozenges mixed with “ … mannitol. And mannitol is one of the things mentioned by some studies as blocking zinc’s action.” Anyway, the published evidence is conflicted.] Apparently, zinc is toxic to bacteria, and useful for the immune system. Anecdotally, I’ve been taking them for 2 years whenever I get a cold, and they have improved things significantly for me. I think they’re even more effective when used preventatively before symptoms start.

In my adult life, I’ve witnessed two waves of mass media reporting on the science of income and happiness. The first wave of reporting, from around 10-15 years ago, claimed that income is important for happiness up to about $75,000 — after that, it stops making people happy. For example:

That’s not an isolated headline. Here are others:

Those headlines were all based on one 2010 study by Daniel Kahneman.

After the above headlines came out, in the last several years, there’s been a shift in the reporting — now, we hear that there’s no limit to how much money makes people happier:

Those are based on a newer study by Matthew Killingsworth, which looks at moment-to-moment happiness; basically it uses data from an app that asks people their income, what activity they are doing, and how happy they are, at any given moment.

So, why all the ping-ponging in the media reports? What’s the truth?

The saga of the research turns out to be “exciting” in that it encapsulates incompetence in both academia and the pop-science news media, but it also shows how professors and reporters were, admirably, willing to correct themselves. But, despite the correction, it hasn’t led the media to give readers a much more accurate picture.

Issues with the first study

To solve the conflicting headlines/studies above, there was an adversarial collaboration between Kahneman, the professor behind the first study that sparked media attention, and Killingsworth, who did the second study. That means they wrote each other back and forth arguing why their study was right, in the presence of a mediator, until they got to the bottom of the differences.

Through that process, they determined that the first study had big issues.

Issue 1: $75,000 was not actually the cutoff, even accepting the original study’s methods

Although the media reported $75,000 as if were the magic number, the adversarial collaboration write-up notes: The threshold of $75,000, which has been frequently quoted, is simply the midpoint of the “60 to 90K” income category.

Also, the 2010 study actually found that the benefits of income merely slowed around $60,000 - 90,000, and only stopped around $90,000 - $120,000. Here’s the graph from the adversarial collaboration study, showing the problem:

The graph makes clear that money only fully stops contributing to happiness somewhere between $90,000 and $120,000.

So the “$75,000” claim was always an artifact of a lazy academic abstract, and lazy journalism on top of that.

Also, the first study was from 2010, and we’ve had a lot of inflation since then. So “$90,000 - $120,000” means $130,000 - $174,000 in today’s dollars.

Issue 2: A “ceiling effect” made it so the study couldn’t measure happiness beyond a fairly low level

Kahneman’s scale of happiness was not sufficient to actually measure happiness; instead, it was measuring something more like “not being unhappy” — and that’s why they couldn’t pick up further gains after a certain income level.

The collaboration authors, Killingsworth and Kahneman, write:

In the range of high incomes, in particular, the average reported positive affect is 89% of a perfect score … The high density of maximum scores … indicates that the items do not adequately discriminate among degrees of happiness—there is a ceiling effect.

A more precise summary, they say, would have been:

the natural summary would be “Unhappiness diminishes with increasing income, but there is no further progress beyond ~$75,000 [$90,000 - $120,000]”

So unhappiness doesn’t decrease beyond that point, because the kind of unhappiness caused by material struggle has mostly been solved. But the second study picked up on greater degrees of happiness, and found that does keep increasing with wealth, without a cutoff.1

Subscribe now

Income increases moment-to-moment happiness without limit, but the effect is TINY

The reason I dug into all this was that I read this article by Spencer Greenberg and Amber Dawn Ace, which noted how tiny the impact was on happiness. Could that be right? So I read through the original studies to verify it.

The adversarial collaboration mentioned above says that the effect is tiny in one paragraph in the middle:

It is important to note that the relationship is weak, even if statistically robust. The correlation between average happiness and log(income) is 0.09 in the experience sampling data … and the difference between the medians of happiness at household incomes of $15,000 and $250,000 is about five points on a 100-point scale. The flattening and accelerating patterns are even smaller modulations of a small effect.

All those headlines in the world’s biggest newspapers, over merely “smaller modulations of a small effect.”

The concrete numbers are stark. Killingsworth and Kahneman conclude that increasing income from $15,000 to $250,000 gives you just a 5 point shift in a 100-point scale. [Side note: I downloaded Killingsworth and Kahneman’s data and tried to replicate their quote above, and found that the professors mixed up the median and the average; the above quote from their paper actually refers to averages, not medians.2 The error doesn’t impact their bigger point.]

What does the small effect look like? Greenberg and Ace also downloaded Killingsworth and Kahneman’s data, and made these informative graphs from it:

So happiness at any given moment does rise with income, but the effect is small.

Few external conditions change moment-to-moment happiness much!

The effect of income on moment-to-moment happiness is tiny, but, as Killingsworth and Kahneman note:

the emotional effects of other circumstances are also small … the effect of an approximately four- fold difference in income [for example, going from $62.5k to $250k] is about … twice as large as the effect of being married, about equal to the effect of a weekend, and less than a third as large as the effect of a headache.

Quadrupling your income is about as impactful as the day being a weekend, vs a weekday. Is that plausible?

I have an app that I use called Dalio, where every day it pings me once, and I click a face to tell it how happy I am. I use it because I’m a data wonk and find these things interesting.

I ran the numbers, and in 2023, on a hundred-point scale where the faces above are worth (100, 75, 50, 25, and 0, respectively,) my weekends were 2.4 points higher than weekdays.3

That’s in line with people in the Killingsworth dataset, in which the median person making $65,000/year scored 62.9, while the median person making $250,000/year got 65 (a 2.1 point difference.) Per the study, that 2.1 point difference is “about equal to the effect of a weekend.”

The study, and what I use, are different-but-similar apps, and not precisely comparable. But the point is that my personal data show a similarly small effect from a weekend. So I think their study is plausible.

Their study also suggests that being married is worth about 1 point out of 100 (though that combines both unhappy and happy marriages; probably there are big effects in either direction, which average out to a 1-point effect for the whole married population.)

Killingsworth and Kahneman also find that a headache is worth about negative 6 points — enough to put even a married rich person on a weekend back in the red.

Fortunately, most headaches are relatively temporary.

What does increase moment-to-moment happiness?

While changes big external things has relatively small impacts on happiness at any given moment, the activities one engages in makes a big difference while a person does them.

Here’s a UK study from 2015, using essentially the same method:

I don’t think anything is too surprising that ranking, but it’s interesting and useful to see it quantified.

Income has more of an impact on life satisfaction

The above analyses consider happiness measured at any given moment in life. That does seem like a reasonable measure of happiness — aren’t our lives in some sense the sum of all our moments?

But it’s not the only reasonable way to measure it.

Spencer Greenberg puts it eloquently:

Are people bad at remembering their day-to-day emotions, so that when they answer life satisfaction questions, they overestimate or underestimate how good their life is? Perhaps. But more plausibly, life satisfaction and hedonic wellbeing are different elements of happiness, rather than two different ways to measure a single, unitary trait. Most people value positive emotions, but they also value other things, including career, family, status, material goods, achieving their goals, and money itself. And sometimes, striving to achieve something you aspire to — something that would increase your life satisfaction — will cause more negative emotions in the short run — for example, when you overwork yourself and take on lots of stress to hit a work milestone. 

Making money increases life satisfaction more than it does momentary happiness across a life, but the effect is still modest. Greenberg summarizes the data:

Looking at US data, every doubling in income is associated with … 6 points on a 100-point scale for life satisfaction, and just 1 point for hedonic wellbeing…

So for example, if someone’s income goes from $62,500 to $250,000 (two doublings) someone’s life satisfaction would tend to go from, say, 60 to 72.

That’s an increase, but not enormous compared to the size of the income gain.

Greenberg summarizes:

Making more money is associated with moderately greater life satisfaction 

Making more money is associated with a very small increase in hedonic [moment-to-moment] wellbeing…

In both cases (life satisfaction and hedonic wellbeing), extra income has a greater effect when you start off with less money – or, more precisely, each doubling of income is associated with the same effect.

Share

Conclusion

Some happiness seems to be genetic. A 2016 study in the Journal of Happiness Studies notes:

“Recent meta-analyses of such studies report genetic influences (i.e., heritability) to account for 32–40 % of the variation in overall happiness …

Evolution programmed happiness in people to tell us to keep doing what we’re doing, sadness to advise us to lay low, and discontent to push us to change things up.

Just like people vary on things like extroversion and openness, people also vary plenty on their baseline happiness.

But the above also suggests has people have some control over most of our happiness. Creating active lifestyle habits helps significantly with moment-to-moment happiness, according to studies that track that. Money won’t help much.

Mental habits like mindfulness and gratitude may help. Spencer Greenberg has an interesting website aiming to help people build “happiness habits.” It seems worth checking out.

For “life satisfaction”, money does help, but less than one might think. Presumably it’s mostly a means to an end — it helps people achieve satisfaction by enabling them to start a family, a business, to change the world, etc. Elon Musk, for example, gets the “means to an end” nature of money, and is fully using all his ~$200 billion to change the world in ways that are meaningful to him.

It’s worth noting that almost all of these studies are correlational, and more experiments should be done in the future to test causation.

It’s also true that it’s not healthy to live one’s life always thinking “how do I maximize happiness?!” But it seems worth reading about at least once or twice, to help set big picture priorities and plans.

I also didn’t cover everything in this post. It’s worth noting that studies also consistently find that religion, charity, and socializing (even for introverts) are all particularly helpful for happiness.

Matt Yglesias has a good, recent article on the left-right political spectrum. He argues:

Left and right are not, exactly, two distinct sets of coherent ideas, but they are two different poles of thought — one that emphasizes the existence of a God-given hierarchy that must be upheld and entrenched and one that wants to tear it down in the name of equality and reason.

I agree that “left” and “right” are useful concepts, where the “right” is about preserving hierarchy and the left is about egalitarianism. Hierarchy means that there’s a clear rank order, where some people have power and respect, and others not so much. An egalitarian society is the opposite: socially and economically “flat”, where it’s not so clear which people are more powerful or respected.

I think Yglesias incorrectly puts religion and reason in his definition. Those things just happen to align with the “right” or “left” at certain times and contexts, such as during the French revolution, when the terms were created.

Yglesias goes through the precise history of how the terms were created in the 1787 French Assembly that sparked a revolution in France:

As the Assembly began to gather … they sorted into cliques based on their ideas rather than which estate they had formerly belonged to. The Baron de Gauville, a Second Estate deputy who aligned with the right, recalls:

“We began to recognize each other: those who were loyal to religion and the king took up positions to the right of the chair so as to avoid the shouts, oaths, and indecencies that enjoyed free rein in the opposing camp.”

A random choice of seating still echoes in our language today. The fact that the terms stuck around suggests they are getting at something real.

In this post, I’ll discuss where “small-l” libertarians (aka “classical liberals”)1 fall on this left-right spectrum, which will also reveal things about the “left” and “right” in the process.

Libertarianism is fundamentally different from left and right, but can ally with either

Libertarianism is about an emphasis on individual liberty, which is a very separate concern from caring about either existing hierarchies or egalitarianism.

Because the concerns of Libertarianism are separate, the ideology can find itself allied with either the left or right. Which side changes depending on the time and context. A clear example of how it can change is libertarians’ support for free markets — that’s because free markets are BOTH a force for leveling hierarchies, but also for creating new ones.

The hierarchy created by free markets is the reason that tens of thousands of brilliant people let Elon Musk coordinate their work — whereas under other historical systems, they would instead do the same for King Charles III, or for communal farm director Alexander Lukashenko.

A beautiful thing about market hierarchies is that they are much more fluid and meritocratic than the older aristocratic ones that they replaced. Musk himself is a good example, as he was originally a middle-class, ramen-eating college student who traveled around Canada on busses. (To address a common mis-conception: Musk’s estranged father did own emerald mines at one point, but he lost that money by the time Musk was a teenager. If you previously believed in a simple story that Musk came from wealth, I highly recommend Walter Isaacson’s biography on Musk.)

Market hierarchies are relatively fluid and meritocratic, and the majority of billionaires are self-made, but there are still some Walton and Hilton heirs who didn’t do anything to deserve their wealth. It’ll also take a while for Bezos’s ex-wife to give away her $36 billion.

But even the most merit-based wealth, such as Musk’s, grates on far-leftists. They see that markets have created a new hierarchy, and their egalitarian instincts can’t stand such a steep hierarchy (this is being charitable; the history of socialist countries shows that some leftists, deep down, really just want a hierarchy that they’re on top of. But others are surely true-believers.)

As markets have replaced aristocrats as the primary maker of hierarchies over the last few centuries, leftists have increasingly focused on tearing down market hierarchies, rather than aristocratic ones.

Capitalism and markets are now relatively entrenched in the west, so those of us who think billionaires have a right to their earnings are often seen as “on the right.”

It’s easy to forget that, when markets were new, and monarchs dominated, people with the same libertarian, pro-market values were aligned with the left!

Case Study of a Libertarian in the French Revolution

The street I grew up on, and a street I lived on in college, and the street my PO Box is on, are all named after the Marquis De Lafayette, a French noble who played a critical role in winning the American Revolutionary War.

He was a true believer in the libertarian enlightenment ideals the American founders had, and he hoped to bring them to France. When Lafayette served in the French National Assembly, he sat on the left.2

In particular, a major early decision was whether the clergy, nobles, and commoners should each get one vote as a faction (which would favor the monarchy, because nobles and clergy both supported it) or whether every individual representative should get one vote (in which case, the commoners would have control, due to their higher number of representatives.) Lafayette, although a noble himself, broke ranks, and voted for the more democratic option.

Encyclopedia Britannica describes:

Lafayette supported the maneuvers by which the bourgeois deputies of the Third Estate gained control of the Estates-General and converted it into a revolutionary National Assembly.

With Jefferson’s help, he composed a draft of the Declaration of the Rights of Man and of the Citizen, which he presented to the Assembly on July 11.

Here’s what was in the declaration, again per Britannica:

The basic principle of the Declaration was that all “men are born and remain free and equal in rights” (Article 1), which were specified as the rights of liberty, private property, the inviolability of the person, and resistance to oppression (Article 2). All citizens were equal before the law and were to have the right to participate in legislation directly or indirectly (Article 6); no one was to be arrested without a judicial order (Article 7). Freedom of religion (Article 10) and freedom of speech (Article 11) were safeguarded within the bounds of public “order” and “law.” The document reflects the interests of the elites who wrote it: property was given the status of an inviolable right, which could be taken by the state only if an indemnity were given (Article 17); offices and position were opened to all citizens (Article 6).

I italicized the editorializing by some anonymous Britannica editor, who thought it worth opining that property rights reflect “the interests of the elites.” That just tells me that the editor is more sensitive to hierarchies than to the pro-liberty enlightenment principles that Lafayette and the founders cared about.

Unfortunately for Lafayette, the population of Paris had values more like those of the Britannica editor:

… the day after a mob stormed the Bastille, Lafayette had been elected commander of the newly formed national guard of Paris. His troops saved Louis XVI and Queen Marie-Antoinette from the fury of a crowd that invaded Versailles on October 6, and he then escorted the royal family to Paris, where they became hostages of the revolution.

For the next year, Lafayette’s popularity and influence were at their height. He supported measures that transferred power from the aristocracy to the bourgeoisie, but he feared that further democratization would encourage the lower classes to attack property rights. When a crowd of petitioners gathered on the Champ de Mars in Paris (July 17, 1791) to demand the abdication of the king, Lafayette’s guards opened fire, killing or wounding about 50 demonstrators. The incident greatly damaged his popularity, and in October he resigned from the guard.

When lunatics like Robespierre and the Jacobins started to get control and began mass-murdering people, Lafayette called for volunteers to put them down by force. When only a few supporters came forward, Lafayette tried to flee to the Netherlands, from where he planned to escape to the US.

However, he was arrested on his way by Prussian troops for his role in bringing down the French Monarchy. Wikipedia describes:

Frederick William II of Prussia … had once received Lafayette [as a fellow noble], but that was before the French Revolution—the king now saw him as a dangerous fomenter of rebellion, to be interned [imprisoned] to prevent him from overthrowing other monarchies.

The response of officials in the 5-year-old United States of America is kind of fascinating3 but in the end, Lafayette spent 5 years in various jails.

While Lafayette and his immediate family survived the Reign of Terror, members of his wife’s family did not. He then retired from public life, partially to protest the dictatorial regime of Napoleon Bonaparte, and settled down as a gentleman farmer.

One might be tempted to think of Lafayette as a “centrist” — and that’s true, if one only looks at the right-left (hierarchy-egalitarian) dimension.

But it’s more accurate to see that Lafayette (and the American founders) were not “centrists” but rather what would today be seen as small-l libertarians, or classical liberals, who saw that individual freedom, markets, and private property were the keys to a new, better, freer, system.

Subscribe now

The left-right binary leaves libertarians out

The right-left (hierarchy-egalitarian) dimension is useful, and the primary axis of modern politics, and I think Yglesias does us a service by spelling it out in clear terms. But the libertarian dimension is also real and relevant.

How to define the core of the libertarian dimension?

It’s something like: individual liberty vs collective control. In other words, does society tend to trust individuals, or the collective will, via government?

We can imagine one axis that’s hierarchy vs egalitarianism, and another axis which is individualism vs collectivism.

Those sound a bit similar, but they’re different: The first is about how “flat” or “steep” the society is, while the second is about whether society trusts individuals, or whether it holds them to the judgement of a broader collective, as expressed via government.

All societies have some regulations that prioritize the collective will over individual will, for example, regulations on medicines, on gambling, on starting a business, etc. The “will of the people” currently determines whether you can start a business offering to sell a new kind of medicine. It determines whether you can pave over a swamp and build a housing complex instead.

All libertarians want less regulation than we have, and the most radical want to end all regulation — they are purists on the individualism vs collectivism spectrum.

Other ideologies are the opposite, often seeing everything through the lens of collectives. Socialists like to talk about “the workers” vs “the millionaires and billionaires” (Bernie’s favorite phrase, though he started saying just “billionaires” after he himself became a millionaire.) Woke people like to focus on racial and gender groups. Nationalists focus on countrymen vs foreigners.

Libertarians, in contrast, are generally baffled (I am, for the record) by the idea of feeling guilty for being rich, or being a certain race or gender. Libertarians tend to distinguish less between nationalities. The important thing is being a free individual.

So a leftist looks at Elon Musk as sees “a billionaire” who needs some leveling, and a French-Revolution-era rightist would look at Musk and see a boorish, uppity peasant. They’re both seeing people as representatives of groups. In contrast, an individualist (libertarian) looks at Musk and considers him on those terms.

Share

Conclusion — “left” and “right” are valid concepts, but other dimensions are important too

Politics could be described using just the right-left axis. Or that, plus the individual liberty - collective control axis. There are surely many more spectrums one could add; the more you add, the more precisely you can describe someone’s political philosophy, but the longer your description.

Using just the two axes above, we’d get the following 4 quadrants:4

Individualist, pro-existing-hierarchy — Think of Elon Musk and Ayn Rand fans.

Individualist, egalitarian — Think of center-right people in Nordic countries, and moderate libertarians in the US.

Collectivist, egalitarian — Democratic Party loyalists, socialists, etc.

Collectivist, pro-existing-hierarchy — Republican Party loyalists, conservative religious people, etc.

Practically, most people are focused on hierarchy-vs-egalitarian battles, which is why discussion focuses on right-vs-left. But several percent of people are more focused on individual liberty, and it’s a bit relevant for political parties who they ally with, if anyone. If the state of the Libertarian Party is any indication, libertarians may be aligning more with the right (they’re also getting labeled as “right-wing” by the left more and more over time.)

Anyway, I think Yglesias’s write-up is great. I just also want to remind people of the axis that libertarians (and I, myself) care more about. I suspect it is a better way to understand various influential people, whether that’s historical influencers like Lafayette, Adam Smith, and Milton Friedman — or modern day players like Sen. Rand Paul and the George Mason University economics department. At least a couple percent of the population are better explained by this axis than the standard right left one!

Share

Share

Subscribe now