Vassar (oft Columbia) behavioral economist (formerly White House & Cornell) working on trust, inequality, climate change Author of https://xitter.cf/t.co/ApYDXEtNdJ

New York
Joined February 2009
Time to do a proper 🧵of my book whytrustmatters.com. I only started twitter when it came out, now after a year of experience, using the occasion of the paperback release to try again. Also a year of book tour talks have helped refine my sales pitch! Why Trust Matters - 1/n
3
10
1
49
another great smbc for econ 101 classes smbc-comics.com/comic/hun
This wasn't my most popular comic, but this panel makes me laugh every time I read it.
1
170
On bottlenecks: 1) economists have been exactly right about little extra growth despite very fast tech change. They thought a group of econs I organized were crazy >2 yrs ago at the AI labs when we said "1pp extra growth/yr through 2030". 2) bottlenecks *still* not understood 1/8
14
97
16
683
105,187
Reacting to the comments on my centibillionaires tweet. “The state is evil. Taxes are theft. Revenues are wasted. I am a communist (not to mention that I am french)” I get the message.😟 But I am still struck at the intensity of feelings (insults?). Really? That bad? Nothing good to say about the state? Fine that centibillionaires pay proportionally less than some of you? The terrain has too many mines (should probably replace “mines” by “drones”) for a discussion to be productive. But I cannot resist making a few points. If the firm has no liquidity, the tax could still be paid by giving the state non-voting shares. If the firm does well, the government will benefit. If it does poorly, so will the government. Sounds communist to you? This is exactly what the profit tax does in every country in the world: the state gets a lot when times are good, nothing when times are bad. The state is already a passive investor. Yes, the funds are not idle, and much is invested in new projects (I knew that, did not need to be told). But if the projects make economic sense, which I assume most do, then the wealth tax should have little effect on investment. The wealth tax does not affect future returns. If they need more funds, Musk and all should have little trouble finding investors. Happy to have a polite discussion.
74
42
9
234
31,609
Just such a fun chart here from @greg_ip. If only we could tease out some meaning. Ah, well, nevertheless. 🤣
11
108
22
514
109,589
I'm thrilled to be launching Econvitals today. It's been the main way in which I have been tracking and understanding the economy over the last several months. And I decided I might as well share it with all of you--for free. Has a combo of website, email lists & social media.
45
212
50
1,472
238,855
Today: California billionaire wealth tax prop 40. Throwback: One of the papers I most enjoyed writing some 15 years ago (with Farhi and later Sleet, Yeltekin) was on Wealth taxation, nonlinear in a political economy with no commitment. Let me explain what we did. I am not trying to take political sides really or advocate one thing or another. Enough of that and it is good to think of the economic logic in a basic way. As you see, the paper is more subtle. The paper modeled a situation where you (e.g., voters, on average) want to do a "one time" wealth tax and promise never to do it again. The incentive to do so is increasing in the amount of wealth inequality. You have nonlinear taxes on wealth and income, so you have rich tools. However, you can't bind future politics or policies. Every now and then you reevaluate (vote). Sounds familiar? Indeed, various basics of the real world tradeoffs are present. Obvious: this is a time-consistency problem that leads to a 3rd best (third world?) outcome (not your standard optimal tax one, where you assume you can commit). Well known and studied. At its most basic, the temptation is confiscation (slogans: robin hood, socialism), but the expectation of this happening in future is of course a problem (slogans: capitalism, property rights). Less obvious: what it is the best you can do today if you with your tax tools, given the situation and constraints. AGAIN: you are not allowed to say "I promise never to tax again". ALSO NOTE: the options are also not necessarily just a choice between taxing wealth now or never doing so. (As the debate on this oversimplify.) Logically: taxing never may just not be feasible/credible. Equally important: you may come up with a different more intermediate plan. Let's see... Indeed, the model spits out something subtle. You actually declare a more moderate but progressive nonlinear tax on wealth that will be in place permanently. It is progressive with rising marginal tax rates: indeed, you subsidize savings for the poor (!) but tax the rich. In fact, you want to promise this tax today for the future, even if inequality is zero today. Why? The credibility problem depends on the amount of wealth inequality. That is what affects the incentive to renege on promises and tax away wealth. The solution tries to deal with this by creating an environment that will lower or keep inequality in check, to avoid a temptation to fall into the time inconsistency revision. A progressive tax goes in that direction both mechanically and in terms of saving incentives. (BTW, the fact that the tax comes out progressive is very clean and surprising, given that the optimal tax literature rarely obtains such results.) Insight: By keeping inequality in check, it makes policy more credible, avoiding falling into a much worse 3rd best. At the time, we called it a "compassionate conservative" solution: even if you want zero taxes, you compromise and do some progressive taxation, to avoid more radical taxes later. Metaphorically, the best way to deal with a tide is not by ignoring it, denying it or swimming frantically, you may have to float with it, adapt and swim moderately finding the right moves. Links: (older ungated): crei.cat/wp-content/uploads/… published: academic.oup.com/restud/arti… Footnote: We titled it capital taxation, but in reality it is more exactly about wealth taxation, e.g. it holds even in a world without capital, where people hold assets against each other etc.
6
70
12
305
33,544
By every measure, Opus 4.8 beats human experts as peer reviewers in a study of 124 Harvard social science papers covering a range from statistical analyses to formal models to qualitative work. As academics, we need to figure out how to adapt the practice of science to AI, fast.
258
261
90
1,709
103,944
The problem of AI for academic writing, and why you don't notice it in your own AI writing: AI loves to use vague language. It will produce something that’s consistent with what you want it to say. The problem is that it will also be consistent with other interpretations that you didn’t intend. This confuses readers: which interpretation was meant? But you won’t notice the problem when you proofread your AI’s output. That's because you’re checking whether the output is consistent with what you meant (which it will be), without thinking about how it could be misunderstood (which we're usually unskilled at, due to the curse of knowledge)
19
23
9
222
28,247
Ben Ho retweeted
🧵New post in my AI series: Can a million parallel agents create superintelligence? TLDR: More agents buy speed, but intelligence gains diminish fast, and much faster than RSI models. Thanks @noahpinion for the encouragement! 1/25
19
48
15
303
142,229
I made a list of 100+ free social science datasets on GitHub. Check it out, and if your favorite dataset is missing, submit a pull request or reply! github.com/toddrjones/datase…
4
128
2
489
18,324
The way you pay for the welfare state is by imposing taxes on the middle class
54
273
53
2,724
399,452
Launched today: @ember_energy has EV sales data. 56 countries, and calculations of energy demand impacts ember-energy.org/latest-insi…
9
27
10
111
8,068
I got a few questions about the Maryland in class AI study and how it relates to @alldaytahelps. Short answer: you need care to design AI assistants! A few things I notice (and we see these bad patterns *all the time* in university-created in-class AI): 1/10
3
6
1
28
3,921
My college classmate Matt and I just used AI to create a Erdos #374 solution w/ a Lean-verified proof. We used a reciprocal-prime equidistribution estimate, Harman’s almost-all short-interval theorem, Weil’s bound, and the analytic large sieve. Full proof: papers.ssrn.com/sol3/papers.…
Erdős Problem #374 is solved! Jonathan Hartley and Matthew Olson proved the Erdős–Graham conjecture that D₆, the set of numbers needing six factorials to form a square, has positive density. They formalized the complete solution in Lean, and we verified it. erdosproblems.ai/374
17
22
3
188
14,088
I'm a fan of Japanese zoning; but I think these Japan-US comparisons miss the nature of Japanese income stagnation Tokyo rent-to-income burdens are comparable to the US; space is often cramped; commute lengths are longer than NYC on averagee thedailyeconomy.org/article/…
14
13
4
153
12,239
one example of Bob Hall as "intellectual butterfly" I was lucky enough to have him as chair of my dissertation oral exam committee on behavioral econ and how people apologize. he had thoughtful and deep questions on the origins of apologies. congrats to both and well deserved!
A quick reaction. Both well deserving of the prize. Research styles could not be more different. Woodford is a cathedral builder, building stone by stone. His main book is an encyclopedia. Hall is an intellectual butterfly, looking for new angles. His main contribution is a superb insight, which changed the way we think about consumption. It takes both to make a field advance. Congratulations.
1
28
1,531
Human intelligence is also spiky.
This is a slide from Terence Tao's 'Math 2.0' public lecture yesterday evening at Caltech. I am highly confident that, if the cure actually worked, almost any Stage 4 cancer patient would tell you that human comprehension of its mechanism is completely irrelevant to them.
4
3
41
2,592
1/ Just turned 74 years of archival National Labor Relations Board annual reports (1936–2009) into machine-readable tables.~2,800 tables, each with a CSV download and the source scans, plus bulk Parquet. Read by Claude (@AnthropicAI). nlrb.andrewjoung.com
1
1
5
117