Review
Re: QB sneaks, the obvious tradeoff that's not factored into the naive model is the risk of an injury to the QB. Running the QB can be worth it if the expected points added are high enough (goal line or pivotal 4th and 1, or if your QB can often pick up 10+ yards on a draw), but I doubt you want to roll those dice on a midfield 2nd and 4 sneak in the first quarter of a regular-season game.
(God, American football is a beautiful game. I miss the days when I could enjoy it because I didn't know about the CTE.)
Great digest, as always. My favorite parts were the link to the US census policy explanation and the reminder that most people don't distinguish between choices and mandates.
An idea I think is cute in theory but would be tricky in practice:
How about just adding social friction to better allocate scarce resources?
Charge high ticket prices, but use the proceeds to run a scholarship program so that anyone motivated enough to submit a decent application can get in for free.
It’s that time again. Here’s all the news and items of interest that isn’t something else.
Things that didn’t make it in to try and stand them on their own include whether America or France has better food and various claims related to Y-Combinator, in addition to the usual additional categories of energy, childhood, fertility and housing.
Bad News
All publicity is good publicity, act accordingly and solve for the equilibrium edition.
WeWork may not do so much longer, substantial doubt about staying in business.
37 year old man has $15k in savings, owns his van outright, but has no skills, works as a dishwasher and has no idea what to do with his life. Nothing brings him joy, he is never happy. What to do? I don’t know, I raise it more as an interesting question. What would you do in that spot?
Yes, some people will choose to be homeless and likely do drugs even under what seem to you like very bad conditions, even if they could choose to get a job instead. One of the tradeoffs involved in the issue is that some people really do prefer this path, or really do now prefer it once the drugs got their hooks into them, if given the option. So if we subsidize them enough they’ll do that, size of that subsidy matters and the supply curve slopes how you think it does. Which then must be balanced against humane considerations and offering help to those who need it, including who need it to recover.
In case you missed it and self-recommending: Patrick McKenzie on credit card debt collection. As terrible as it seemed likely to be, it is worse, with the feature that if you know the right words to say payment is optional. Sarah Constantin offers her thoughts in this thread.
Rich people often do not consider themselves rich, or at least do not feel rich. The part I do not understand, no matter how many times it is explained and how much I get the gears involved, is the vast array of people who make $150k+ per year and have no savings, who would adjust to and spend any amount that comes in.
Robin Hanson notices that often people use famous thinkers to construct personas that represent the thinker, then use those personas to extract far more opinions, and this often dominates arguments and argument details in people’s heads and thus in impact, to extent that Hanson says the persona is the primary way thinkers impact the world. I think that goes too far. His example of Adam Smith is illustrative. There is much talk about the wrong question of whether he was (or would have been, I guess?) left wing or right wing, and people do indeed mostly ignore his argument details. I read The Wealth of Nations and was impressed including the digression on silver but mostly don’t remember the details either. What matters is mostly, in my model, the core idea that gets passed along, in this case largely the invisible hand.
A reminder that if someone thinks having a need justifies the use of the force (via a law or otherwise) to satisfy that need, they may interpret a claim of need as a justification of violence. Also a reminder that when you write about certain topics let us say you might not want to read the comments.
States are increasingly banning online pornography, which they technically describe as ‘age verification laws.’ You can guess how excited people are to give their identification to a porn site, so this is effectively very close to a ban, which can of course be evaded with a VPN for those who actually care. We will see if this blatantly unconstitutional move is upheld, in the meantime sites like PornHub are pulling out of entire states. Then again, this might be the first time states moved to ensure our kids actually learned something important about modern technology, and putting a trivial inconvenience in front of porn has several advantages. It needs to feel special.
Is There a Reliability Crisis in America?
I do worry this kind of thing seems to be more common, but this could easily be a case of rising expectations, cherry picked issues and not properly remembering the past. One note especially is that in the past there were essentially no review systems and to get something done you often had to know a guy who knows a guy, whereas now no one knows a guy who knows a guy, you’d be lucky to know a guy at all.
Real estate people in my experience run the gamut. A great one not only knows how to keep appointments, they know how to ensure the other agents mostly also keep their appointments and their help and guidance is worth every penny. A random one you call because you want to see their property? Could go any number of ways.
Our experience with sitters reinforces this point. Many we attempt to hire will cancel last minute, with a shockingly high rate of sudden family health emergencies, the good news is the internet often makes it possible to scramble for a replacement. I also note that, in a similar way to Uber, we have learned a different kind of trust and reliability that we place on strangers now that we would have been called insane for doing in my youth.
How reliable are appointments and people more generally? If chosen at random without a verification method, not all that reliable and it seems worse than in the past. However, we also have various rather robust rating and review systems in place, which both identify and motivate good behavior quite well in many cases.
Small Problems With Big Tech
Zoom’s terms of service include the following:
Google Meet exists, is about the same level of free, can even be HIPAA compliant, and works fine.
Facebook non-binding oversight board is not binding, and is ignored once again. Remember that it is better to have the board and ignore them than not have them, whereas our instinct is to use the board as justification for additional blame. Resist.
Twitter will be getting rid of the “Tweepcred” system entirely real soon now, and you will no longer be effectively penalized for interaction with low-reputation accounts or not having enough followers. The claim is that the old system often crippled the reach of small accounts that had done nothing wrong. The issue is that this is part of the bot and spam defense system, and that system is already in deep trouble, so perhaps now is not the time to weaken it further and thus we are waiting to implement this change.
Google Still Planning to Permanently Delete Inactive Accounts
I’ve mentioned it before but yes, the Google two-year inactivity threshold is really bad. In particular, what happens if you get thrown in jail?
The issue above was already present before. This only makes it worse. An obvious response is ‘don’t have a single point of failure’ but easier said than widely implemented. If my Google account got deleted or permanently stolen, I don’t know exactly how bad it would be, but it would be very bad. Outright deletions will be far worse than current nightmares, and more common. Yes, you can say ‘tell your lawyer or a family member or friend your login information and have them login for you every year’ but a lot of people will neglect to do this, or perhaps not be able to, or someone will forget.
I strongly urge Google to reconsider the account deletion policy. The amount of bad will and paranoia it will generate will be astounding. The costs of data preservation will be dwarfed many times over even in the best case, and who would ever throw data out at this point anyway?
Crypto Might Not Be Entirely Safe
There is no such thing as a safe way to store your crypto. If you have a physical wallet, perhaps you end up with a bankruptcy filing that refers to “The Wallet Event” where you lost the password to $38.9 million worth of coins. Every solution that guards against ‘losing the physical key’ or ‘the person who knows the password is hit by a bus’ means duplicating the information. There is other risk in storing your crypto with someone else who will handle this for you. No safe harbor. Ask which path is actually safer for you, and remember that every time crypto is permanently lost that means more value for the rest of us.
There are also some new proposed US tax regulations, that many claim would effectively force a large portion of the crypto ecosystem to geoblock the USA.
As I understand it, the new rules say that the requirement that brokers provide 1099s and ensure proper withholding on sales should apply to sales of all digital assets, and would treat essentially all crypto apps that allow such sales as brokers with that obligation.
Also as I understand it, this is what happened.
Yes. This is a failure to understand the technology, where ‘the technology’ refers to the United States Government.
You say that it is impossible to comply with the law? Perhaps you should have thought of that before you designed a technology incompatible with the law. Instead, you insist that it is the law that must change. You are fighting the law. Guess who wins.
Either find a way to give the taxman his cut, the one he says he is entitled to verified the way he wants, not the one you think is fair whenever you feel like it, or otherwise not violate their rules, and you can have your technology. Otherwise, I am guessing, not so much.
It would be a real shame if the same thing happened to AI because of some issue like copyright, professional licensing or liability. A real shame.
Government Working
We finally have a candidate we can all support: Alexander ‘24. He’s the great.
Unshackled: A practical guide for highly-skilled immigrants to thrive in the United States. Link goes to the book’s website.
Yes, we need much higher prices to see the Mona Lisa.
On our recent family vacation to our nation’s capital, we noticed that all the public attractions cost zero dollars, so we had no access. The deficit keeps growing.
How goes the Chips Act? Working as designed. Unions work to block plan to bring in Taiwanese workers to jumpstart the Arizona TSMC plant that is sitting idle for lack of workers with the necessary core skills. Relations between TSMC and the union are degenerating fast, the union claiming TSMC is merely trying to import cheap labor and lobbying to deny the workers visas, and the Taiwanese firing back calling Americans lazy. The plant may never open.
Justice Department sues SpaceX for focusing its hiring on U.S. citizens and lawful permanent residents, despite the department of defense requiring that if they are to work on its projects SpaceX can only hire U.S citizens and lawful permanent residents. Also you need to match those requirements to work at the Department of Justice, even when completely unrelated to any reasonable security concern.
The whole legal situation is ludicrous in general. The solution is, obviously, to clarify that the security rules take precedence, and discrimination is legal to the extent it is required by the practical needs of the security rules.
In the case of SpaceX, it seems impossible for this not to be grandstanding and political targeting by the Department of Justice. Rockets are a place we actually care about operational security, no reasonable and neutral person would be suing SpaceX here.
They totally wouldn’t be replaced by a new tiny cabal of corrupt and incompetent politicians, right? There’s no way the people inevitably vote for such madness.
Freyja is correct.
This form is ordering you to lie. You lie your ass off. There are zero people in the world, who are both capable of getting such an appointment, and also lack any association with any person, group or business venture that could be used, even unfairly, to impugn or attack their character. Yet quite obviously the goal is for you to click ‘no.’
This is an epidemic pattern on all sorts of forms, drilling dishonesty into the people.
Government Planning
A strong case is made for dynamic budgetary scoring, illustrated by the case of high skill immigration. It would cost $3.1 billion over 10 years to allow foreign STEM graduates to get green cards because of administrative costs. Everyone agrees this would have massively greater economic benefits and growing the population, net earning the government money. But the CBO is instructed to score such actions only as costs, so it becomes difficult to implement when everyone has a fixed budget to spend.
There are two distinct problems here.
The first is that we have grown so rhetorically and procedurally and even legally attached to the budget scoring of proposals. Our actual laws get massively warped based on what will technically score what number when the CBO uses its procedure on a ten year window. We get rules that change dramatically in year eight or nine to make the numbers work. We get ‘annual fixes’ no one wants to pay for until the last minute. We get things far less sane than that. We also get one set of actions where we need to find ‘pay-for’ things that technically make everything add to $0, and other sets of actions that don’t follow this process and so spend tons.
The second problem is that once we start with dynamic scoring everyone will argue about and abuse the hell out of it, so we cannot have nice things. I see the narrow argument for ‘if this literally means more people then we get to count those people’s new tax revenue’ but those people also bring costs, the magnitudes are difficult to get right, and once you open that door why aren’t we dynamically scoring everything else?
I presume the right answer is that we should have two or more scores. There should be a no-dynamic-adjustment number, then a only-the-agreed-upon-adjustments number, then we can have a fight over the loony toons numbers everyone gets when they use some mix of maximally wishful thinking and trying to get the real answer.
An alternative solution is to treat this as an incentive problem rather than a calculation. You do not want the technically right answer. What you actually want is a system where you don’t kill clearly net profitable proposals, so rather than make a dynamic adjustment of uncertain magnitude, use a trigger where sufficiently justified items whose budget allocation then gets determined by the CBO or a similar body can be scored as net costing exactly $0, and existing outside the budget calculation. The reason we stick to a budget is to not go nuts with spending, not because we only have so many dollars.
Vivek Ramaswamy proposes immigration policy of only letting in those who can show they will be a net economic benefit to America, and that they can pass a civics test. Worker group fires back that there are over a billion people who could meet this requirement, Bryan Caplan points out this is not exactly a case against such a policy.
Trump is proposing a 10% universal tariff. Krugman projects this would directly cost <1% of GDP, because as a uniform tariff it avoids ever being truly terrifying. The proposal is terrible not for the direct impact (it hurts us to do this, but not that much) rather it is terrible because it generates a free-for-all rather than ending there.
South African laws reportedly make it impossible to legally evict people from apartments, so gangs have effectively hijacked (stolen) over 600 buildings in Johannesburg alone. They solved for the equilibrium, although not yet so fully.
As is often the case, the question is not why is this happening, the question is why is more of this not happening.
Together We Fight Crime
San Francisco police commission attempts to forbid police officers from pursuing suspects even on foot. Walk away.
Related: A video of an organized mass robbery of a Nordstrom in California.
As I noted in the comments, we should instead be pleasantly surprised that things are not falling apart more quickly. Once people ‘run the experiment’ and find that such actions are effectively tolerated, why doesn’t the pace accelerate, which in turn makes it that much harder to enforce the law, in a rapid negative feedback loop? Or, alternatively, how long will it take before a new government arises to fill the void left by the current one?
Yale campus police union warns new students in a flier complete with a picture of the grim reaper that the city is crime-ridden and dangerous, telling them to ‘remain on campus,’ ‘avoid public transportation’ and ‘stay off the streets after 8 p.m.’
SBF subsisting on literal bread and water because the prison refuses to provide him with vegan meals.
I presume this is a ‘happens constantly’ thing rather than him getting especially bad treatment? It seems the prison is often ‘described as inhumane’ and no one seems inclined to do anything about it. You can pass all the laws you want, the prisons decide to ignore them, no one cares. Similarly, the court ordered that he be given Adderall to help him assist in his own defense, and the prison decided that was more of a suggestion.
NYPD has 1,500 people in the city – one in about every six thousand in the city – working on ‘intelligence and counterterrorism’? That’s a cinematic level of paranoia and presumed background threat. It’s not impossible that we could both face that level of threat requiring that many people, and also be this good at preventing incidents, but it does seem unlikely.
Want to break up a theft ring that stole your computer? Easy, all you have to do is all the police work, since they won’t do it for you, but also it’s not like it was hard. A story from Brussels (2 minutes).
How Bad Is Crime, Anyway?
Some noble (older) attempts to measure the impact of crime.
Those numbers sound remarkably sane, especially adjusted from 2006 to 2022, in comparison to other estimates. If anything, they’re too sane, $10 million is used as the value of life in some health care calculations. This makes me suspicious. Do people actually have the ability to do these types of calculations? If people will pay more to save one bird than ten birds, what makes us think they can measure the value of 10% of a burglary? What about willingness to pay always involving people underpaying?
The answers here are still too low. I am confident that the correct value of preventing the average prevented murder is well in excess of the value of preventing average medical death, based on how people react to them and guard against them, and the ages of those impacted.
We know these numbers don’t make sense, wrongful death compensation is way too low and pain compensation way too high, hence the ‘back up and roll over them again to make sure they’re dead’ problem.
This seems like a lower bound if you do it technically correctly, causing a temptation to do it wrong, since a lot of the negative impact of crime is the impact on what happens to life in general even when not directly involving crime – the OP also makes this point in its conclusion. A huge percentage of the cost of crime is our worries about and (highly reasonable and often even wise!) obsessions with avoiding and preventing murders, rapes, thefts, frauds and so on. Spaces where one does not have to worry about such things are so much better, often feeling downright sacred, even when the worry doesn’t materialize.
Housing prices feel more accurate in the sense that they reveal true preferences, with the problem that so many things are so tightly correlated with crime rates. When the murder rate is sixty times higher across the county line, so are many other things.
I am guessing this particular methodology is flawed. A particular murder directly lowering prices is acting in part through updates on the expected number of future murders, but if that is true why is the effect so localized? My guess is a lot of this is people acting irrationally, the way people talk about haunted houses and don’t want to live in a house where there was a murder – the whole thing freaks people out, and there’s a kind of blameworthiness or ‘why accept this trouble’ vibe. I’d be more inclined to measure this through changes that alter reasonably expected future crime rates. The obvious candidate as a player of city simulation games is to look at the distribution of police, perhaps, although that also has other effects.
What does this all add up to?
Asking ‘what is the cost of crime’ is both at the same time highly interesting and a wrong or wrongly worded question. What is the alternative world? A world with zero real crime, in the sense that there was no theft, no fraud, no rape, no murder, no assault and so on, no matter how tempting the situation, would look very different from our own. I would happily pay far above 12% of GDP to live in that world, and would expect to end up with higher net GDP very quickly as people took advantage of the new opportunities – it is so much easier to Do Things worth doing, productive things, in that kind of world.
I would also note that I disagree with OP in the percentage of the gains here on violent versus non-violent crime. Knowing you are safe from non-violent crime is a truly phenomenally big game, especially if that is broadly construed and effectively means that you can trust other people’s representations and promises. That is much more valuable than protection against violent crime, unless perhaps if it came with protection from the government as well.
The Quest for Mundane Utility
Because it is built on Vimeo, you can search a Criterion movie for exact words and it will take you straight to that spot.
Wait, you can turn a purse into a backpack this easily (30 second video)?
Can confirm this seems to work.
With the caveat that if you want to be President of the United States, that is not going to work. For most of us, go for it.
If you can both build stuff and get stuff done? Now we’re talking.
Nassim Nicholas Taleb offers his components of True Wealth.
This is a very particular list. Many of these I would actively reverse. In many ways I want Worthy Opponents, not inexperienced enemies. I want walks, not gravel bicycling. I want good water, not foamy coffee. I want tranquility and the lack of a schedule, not no meals alone. Others on this list seems vital and very good.
There are also the things that are conspicuously missing. I would include something like ability to work on important, interesting and challenging problems, ability to follow curiosity, lack of scheduling. There is also nothing here about romance or sex, and especially nothing about family or children. Any list of true wealth without children seems at best highly incomplete.
If you are willing to go into a trade like carpentry or plumbing, you can earn a good wage and be for now safe from automation. Except no one is interested, which is why it is paying so well.
There are jobs that are more intellectually stimulating than the ones on offer here. There are also many jobs for college graduates that are mind-numbing and also that require you to sacrifice your mind while giving you nothing concrete to anchor you to reality and no feeing of accomplishment. Building or fixing physical things lets you see the fruits of your labor, do something obviously helpful, and lets you keep your sanity. You could do so much worse.
Suggestion from Write of Passage that you can use prestige as an anti-indicator of opportunity, in writing and elsewhere. If it is not prestigious, no one wants to do it, so you can get paid handsomely for taking on a low-status project like a romance novel. This is similar to not pursuing the dreams everyone has, like game designer and rock star, if you can help it, although that list also includes writer.
I would however warn that the opposite phenomenon often also applies. There are many circumstances in which high status and high prestige come with high compensation and a better experience, because compensation is primarily socially determined in most places when those involved are not doing business or trade. People strongly think the two must align, value be damned. Thus, one needs to notice the different classes of opportunity. If you write a romance novel and are paid by the sale, you can get paid above your social standing, so that works. If your pay was linked to your job title, that would not work.
Opportunity Knocks
Magnificent Grants offers $100k fellowships to 10 or more people under 25 years old every year, in exchange for dropping out of your current college, school or job, to help them break free of other pressures. Deadline this year is October 31. It is a pure grant, if you are thinking of applying you likely should do so.
Paul Graham wants to fund an eye-tracking startup.
I would worry that the Apple Vision Pro means that eye-tracking is highly advanced or potentially a solved problem already. That helps Paul’s friend if and only if they get their hands on an actual product, but trying to get hold of one early would be the first thing I would try.
Good News, Everyone
LK99 around the world as real science done right, despite LK99 itself turning out not to be a superconductor.
The value of the obvious:
The implication is that what the world needs are successful people noticing which obvious principles they are not so confident in, not merely the ones they are confident are false. It is difficult to say ‘I have no proof or anything but I am suspicious that [X] is not actually true’ for obvious-seeming or standard-banal things, such as my recent statement ‘I suspect mindfulness is a scam,’ but such statements are useful evidence if taken in the spirit offered.
A good example of the obvious from a successful person is Steve Jobs pointing out that asking for help is a superpower, you can ask people for help, including famous and powerful people you have zero connection to, and often you will get it, with very little downside.
It won’t convince anyone that Americans have the best food, but I strongly agree with Patrick McKenzie that the food court is a wonderful innovation, a vast improvement in variety, choice and satisfying a group over having to go to a place in the generic middle. It is almost always fast and cheap, and what hits your spot is often remarkably good at punching above its weight. We are also starting to see actively good places make it into food courts, such as Mighty Quinn’s BBQ or Dos Torros.
Study provides strong evidence that YouTube’s algorithm, rather than leading people down extremist holes, does the opposite. YouTube wants you to watch centrist or left-leaning videos, will grudgingly serve you right-leaning videos, and will require a lot of convincing to give you far-left or far-right content, while being happy to switch back to its preferred affiliations if you show any inclination to do so. If you create a new account, it will be half centrist and half moderate-left with only a small amount of moderate-right. As is the case with AI, they stack the deck in favor of the boring conventional thing, and the conventional people complain that the stacking is not 100% effective.
Claim that Barbie’s marketing succeeded in part because Mattel made it cheap and easy to collaborate on licensed products, as opposed to Hasbro and others who try to extract lots of money and retain careful creative control. As always, when you have a good product, the things you do with it look good, which makes it hard to tell what other things were actually good or would transfer. This was also a much easier opportunity, because the tie-ins were essentially the color pink. You don’t need to be that careful with a color.
Are college towns getting worse? Or is everything else getting better, making college towns less unique?
As far as I can tell this is mostly other places getting better. College towns were living in the future, which is now more evenly distributed. The new unevenly distributed future is online. Meanwhile, college towns are full of NIMBYs, so they stay the same except for higher costs, rather than realizing their potential.
Eigenrobot on the nation’s most extensive attempt at government censorship, of abolitionists calling for the end of slavery. If your response to this is that this is all the more reason why we must censor the bad people, perhaps think harder.
I am not alone, almost half of us each less than three meals in a typical day.
Did the reality meet your expectations? Bryan Caplan investigates this as well.
Some survivorship bias is presumably involved. Still, update in favor of things likely turning out better than you expected on a personal level, and especially that parenthood is likely to exceed your expectations.
Money Stuff
Small businesses less likely to be planning price increases, most of the way back to their previous baseline. Feels like an excellent sanity check on the statistics.
More generally, Tyler Cowen notes that all the economic statistics look very good, which they do with the exception of the deficit and debt. He says the American economy is great, stop worrying about it.
In particular, he says stop worrying about the risk of recession, good times do not inevitably beget bad times except insofar as there will always be occasional bad times. I fully agree with this. Recessions are very hard to predict and the good economic news makes the next one more likely to take longer to arrive.
Where we perhaps disagree is that I do not think the statistics are sufficient evidence that the economy is doing great. Inflation is improving but remains above target even before the need to return to the trend line. Interest rates are high. Both of these have a big practical impact on regular people. Meanwhile, no matter how much economists point out the availability of new or higher quality goods, people continue to not thrive in the ways that most matter. Nor do I think that a large share of savings have ever been well-explained, as the paper linked suggests, by anticipated superior future product offerings. Beware, one might say, the dangers of calculating welfare as the cumulative value of goods, while losing sight of holistic affordances.
Tyler then reiterates his emphasis that congestion pricing should be differentially applied only to residents of an area and not to visitors. It is bizarre to pick this particular scarce good where there is large disutility to overconsumption, and to want to price it lower to those who one thinks get more utility out of it, claiming this improves efficiency. Why is our default not that prices solve exactly this problem? If not, does it not indict the general concept of allocation by price? What other differentiated taxation and pricing regimes based on centrally planned beliefs about available utility would this principle favor?
We have another study claiming that the impact of climate change on world GDP will be minimal.
I agree that we have to do the math and figure out what makes or doesn’t make economic sense to do. And that we are currently both ignoring very good mitigation opportunities, and implementing or considering some highly inefficient mitigation strategies.
However. Bullshit. That size of impact does not make any sense. Even if the GDP calculations are correct and the counterfactual was meaningful, that does not mean human welfare, and this is a case where those two things obviously are vastly different.
Assuming for the moment that AI peters out and the world still mostly exists as normal otherwise, I would much rather live in a world that was 400% of today’s GDP and did not have any further warming, than one with 450% of today’s GDP and had the baseline non-intervention levels of warming, even measured only by direct impact on lived experience of it being too damn hot.
Hell, I would expect the purely psychological impacts of climate change and fear of climate change, including on fertility, to on their own (excluding the actual physical downsides) already have baked in a more than 4% impact on world GDP (to say nothing of mental health or actual human welfare) by 2100, again in the ‘normal AGI-less world’ scenario. One could argue the cognitive effects of high CO2 levels on their own will cost more than 4% of GDP. That does not even touch on the bigger impacts, or the tail risks, or the secondary effects.
How false is the efficient market hypothesis?
The price is set by a combination of all supply and demand. If the price got sufficiently out of whack, those with better knowledge would come in to correct it. They would however need to be compensated enough to make this worth doing. So yes, being good at this will make you money.
Our Price Cheap
Tyler Cowen asks for a general model of why, so often in the world of food, price and quality are negatively correlated.
My model is that restaurants can only mostly care about and focus on one thing. If it is the quality of the food, then they are unlikely to profit maximize and look to charge more, and they will look to find locations and other ways to keep prices down rather than ways to raise prices up. They also will choose circumstances that put them under less pressure to charge more, and less pressure to in turn do the things that justify charging more. Whereas places that are maximizing profits will not focus relentlessly on quality of the food. One could also say that only high quality food will, in many places, let underpriced offerings exist.
This only applies if you control for the places where quality runs in the obvious direction, and have the right baseline. Quality of ingredients matters a lot, if you are not charging enough to afford good ones, or you are outright competing on price on the low end such as fast food places like McDonald’s, or generic places like the ‘99 cents pizza’ shops, then this will catch up to you. Prices that are too low are red flags if you care about quality, as it is another sign that what is being focused on is the price. Chains are using various means to drive down price and value standards and reliability rather than quality, so once they reach industrial scale they generally need to be excluded from this kind of heuristic.
This explanation from a different reader focuses on the demand side rather than the supply side. Demand here has customers who want comfort, who want something fashionable and who want good food, but there is little overlap between them, and those who want good food will often learn to cook it themselves. Unless someone actively seeks good food, quality beyond a baseline threshold is mostly wasted.
In this model, P.F. Chang’s is competing not with the quality family Chinese restaurant, but with Applebee’s, and willingness to pay is mostly about atmosphere not food quality. Thus the negative correlation.
I definitely can see this model. Some days I want the best possible food. Other days I want comfort food more, something I am familiar with and know is good. That still often looks like something out of Tyler’s ethnic dining guide, although not always. Rarely I will care so much about atmosphere beyond a basic threshold, but I’ll still do what I can to maximize food quality, and I have a small rotation of places for such times that I know work well.
The two explanations combine well. Demand is segmented, and supply is also segmented, so it is doubly hard to combine high quality in all areas at once, and the food quality demand curve is lower.
Another note is that over time, if one seeks good food, this changes what one considers a good atmosphere, because your associations build up.
If you want to take advantage of this correlation, you need to control for what things should cost, and know what discount levels are good and which are suspicious.
The Efficient Market Hypothesis is False
Here’s a paper: Do firms bunch at EEOC kink points? Implications for firm value under exogenous risk. Abstract and highlights:
These effects are crazy large. If the number of firms at kink points is 25% higher than expected, that means that 80% of them are not doing it on purpose. If a group that is 80% controls has negative 7.26% annualized returns compared to the control group, well, you do the math.
Also note that this group is real and bigger than one might expect, but in absolute terms it is also small. We are talking about roughly 1% of the relevant employee count ranges (4 spots out of 506), so there is intentional bunching found in maybe 0.2% of firms. There likely are very big selection effects in the intentional group.
(How much employment is lost here? Presumably not so many, since a firm that naturally wants a lot more employees wouldn’t let the cap stop them. It would be reasonable to say on average something like 2 employees if you’re at 14, maybe 5 if you’re at 100 and so on, so maybe these firms are let’s say… 8% under-employing? But with these kinds of returns, that won’t last long. Still, they get replaced by other similar firms. So a snapshot says 0.2% times 8% is 0.00016%, or 1.6bps, or something like 20,000 total jobs in mostly lousy places to work.)
What is most interesting here is that there is a self-identified subgroup of small firms that, when you take out the intermixed controls, has massive, massive negative returns.
Which makes sense. If you are limiting your employee count to stay under a liability threshold, you are telling me several things.
It is not a great leap to think such firms would underperform. It is however a great leap that they’d collapse this quickly in value.
This seems like it would show definitively that some firms are making really bad decisions. It is not best practices or efficient to incur a bunch of discrimination liability. It is not efficient to do the other things this group does, as evidenced by the results. Yet I get people claiming that businesses mostly do the efficient, sensible thing for them, they would never do something like have bullshit jobs on their payroll.
Alas, it is difficult (read: impossible) to short firms with only 500 or fewer employees. What you can do is avoid being long. Thus, if you had a basket of investments that returned market rates, and then you tossed out 1% of it and that 1% was previously underperforming by 7%, you get a modest performance bump. On its own it isn’t much, but you can repeat the process with other attributes or signals.
One can also note that small firms seem to be implied to have negative 7% alpha in general? What’s up with that? Seems like an impossible result. Or at least a highly bold claim. If it was true, then the valuations of such firms would be Obvious Nonsense.
Another symbolic example.
In the Super Bowl, it was clear that the Eagles could call a QB sneak and it would always work. They lined up in a formation that was certain to be a QB sneak. They did the same physical thing every time. It worked every time. Yet in situations where doing this twice or three times in a row would make sense, it is almost never used.
Why? And given that it doesn’t happen in such a competitive situation where the math is so easy, why should we expect anyone else to be doing the right things?
Prediction Market Hypothesis
Problems in prediction markets: What to do if you want to do something if and only if chances of success are above 30%, but if you entrust the market with this decision then the correlation between it being a good idea and you doing it weakens, messing up the market? And how often do you have to f*** around in order to find out when it ends up below 30%? I’d say you shouldn’t take the market as your only evidence, so you’ll sometimes try when it is below 30% or not try when above, and that’s fine?
Manifold markets are reasonably well calibrated overall, but it is still smart to mostly be betting no.
This diverges from other prediction markets, where it is standard for the yes side to have the edge above 40% or so, here you want to be on the no side all the way.
That is especially true for the Free Money markets, that second dot on the left. Those are markets where the outcome is known or all but known, but that won’t formally resolve for a while, so capital costs keep the price above zero. People who have yes shares hold on to them partly for fun or hope, partly in case some no holders need to free up capital. The loan mechanism helps somewhat, but in practice only somewhat.
This is proving a problem in particular for the superconductor markets. The main one is still trading at 8% as I write this, whereas Metaculus is down at a far more reasonable 1%. This is despite the fact that every day I take my loan money and put it back into the market, I now have more than my entire Manifold balance on this. We will see how far that gets to snowball over time.
Modest Proposals
The Fill and Flush method of plane clearing is one of many options all vastly superior to the current highly inefficient method. We are almost maximally bad at getting people off of planes, yet the mild social awkwardness of a systematic method, combined with not wanting to mess with ‘lower row number privilege’ or something, keeps us from changing. I would give substantial priority to any airline that fixed this.
FDA Delenda Est
The FDA requires labels. They are not especially sticklers for accuracy, in some spots.
Requiring such labels seems fine, but if you do that then you need people to be able to rely on them. These are rather obnoxious technical loopholes. I am not mad at those who take advantage, that is the game, but then once it comes to your attention you need to patch the bug.
It is much worse when the FDA does things like regulate sunscreen as a drug, making developing new ones prohibitively expensive, and resulting in us having far inferior sunscreen to other countries. AOC films a video to ask Congress to fix this specific failure, without any sign that she has realized the principle might generalize.
It is still good to notice this in one place, especially given the price to be paid for noticing.
These criticisms seem related to the phenomenon Freddie DeBoer notes here, where he criticizes (from a left perspective) several of AOC specific actions and gets lots of criticism, also from a left perspective, about the symbolism of AOC as a person but no response to his substantive critiques and questions. In both cases, substantive criticism pointing out ineffectiveness of decisions is responded to on only the symbolic, whose-side-is-this-on-or-how-does-this-vibe mostly third and sometimes fourth simulacra levels. Whereas a key reason why AOC and Freddie are worth paying attention to and might accomplish something is that they are capable of, to their great credit, asking what a given physical action might result in, and even in insisting others ask such questions.
I also loved this exchange:
In Medical and Health News
Semaglutide not only causes you to lose weight, it is confirmed that it in turn reduces risk from composite of heart attack, stroke, & cardiovascular death in patients without diabetes but with a history of CVD who are overweight or obese. It is hard to see it could have been otherwise, still good to confirm and magnitude of 20% reduction is big. Results like this will likely make weight loss drugs increasingly standard treatments.
You can now sign up for a human challenge trial for malaria (direct link). They pay $5,400, but you are hopefully not doing this for the money. Amazing that we are getting our act together on this.
News you can’t use without a prescription, about Celecoxib.
25 years ago, they invented a better version of Advil, which targets the same receptor but with fewer side effects. It still requires a prescription for basically no reason.
China had 1.87 million excess deaths in the two months after Covid zero ended, distributed as you would expect. That is something like 0.14% of the population.
Britain to only offer Coid and flu shots to those 65 and older this winter. This seems categorically insane, if you are paying all health costs and not offering free flu shots you are in a downward spiral that I have no idea how to recover from.
Medicine has a modest left digit bias problem, where people place too much emphasis on the first digit of numbers, so things such as being age 39 vs. 40 gets you treated differently. Measured size seems if anything smaller than I would have expected, so that is good news. Heuristics are important, a little inefficiency here seems fine. As post notes it can be used to measure the effects of medical interventions if you can control for other left digit effects.
Estimate from a paper that Covid vaccines saved about 750,000 American lives, most of which would have been impossible without at least some continued mitigation efforts throughout 2020 and early 2021. One can argue either way whether that was enough to justify various mitigations.
How did you do on Covid in terms of preventing deaths as measured by excess mortality? Why did Sweden do exceptionally well? Because mostly only one thing mattered, which was vaccination rates.
And yet we somehow really, really messed up with the vaccine marketing, didn’t we?
Retrospective support for mask mandates remains strong at 63%, whereas only 55% think the vaccine was a critical advancement rather than rushed.
The hint is the next question. What is the approval rate of the vaccine at all? 55%. What is the approval rate of a vaccine mandate? Also 55%.
People literally, flat out, cannot draw a distinction between a choice and a mandate, that which is not forbidden is compulsory and that which is forbidden is not allowed, and that is it.
If nothing but vaccination mattered much for excess mortality, then that suggests strongly that, within the range of realistic responses, the correct answer was otherwise to be as light handed as possible.
Multiple scientists and industry insiders insisted the probability of success for mRNA vaccines be set to zero, because anything else would be wildly miscalibrated. This was not years earlier, this was early on during Covid. Everyone involved must go to Bayes Hell.
The classic story of how ice cream is shown time and again to be healthy and associated with a reduced risk of diabetes, so nutrition scientists first try everything to make the finding go away, making up bizarre stories including reverse causation, then when they not only won’t go away but replicate the scientists say nuh-uh and blatantly lie to us that such effects only belong to foods like yogurt so we can properly suffer for our good health. Every time nutrition science seems like it couldn’t get any more confusing and unhelpful, there you go. I have long since settled into not trusting any of it beyond the supreme basics at most, and trusting my own lived experiences instead.
The reverse causation hypothesis here is that ice cream is known to be bad for you, so if you are in poor health your doctor will talk you out of eating it, so now it only seems healthy. Which is such Obvious Nonsense all around, as is noted in the article you do not see this effect for cake or cookies. If anything, someone in poor health will tend to increase their intake of relatively easy to consume foods like ice cream. When my uncle was dying of ALS they had him eat tons of ice cream to keep up his caloric intake.
My alternative hypothesis is that ice cream is efficient. It is not health food, but it gives an extreme amount of joy and symbolic representation of desert, in exchange for a remarkably low health cost. It is substituting for what would typically be much larger caloric quantities of unhealthy starches like cookies and cakes, while still telling you that you have had your special treat. Your balance of consumption shifts in positive ways. So no, if you had ice cream for lunch it would probably make your health situation worse, but that is not how people actually consume it.
Whereas I am going to go ahead and challenge the causation theories on whole milk versus reduced fat or skim milk. Everyone knows whole milk is superior. So the only reason you would consume reduced fat or skim milk is if you were being sold a bill of goods that it was healthier, and you were willing to actually suffer. Which in turn means you’re probably doing other doctor-recommended things, and especially consuming fewer overall calories.
So the WHO seems to be at least somewhat endorsing… homeopathy? WTF?
Tyler Cowen called his link to this ‘WHO is a fraud’ and, yes, well.
WHO Delenda Est. My lord.
Gamers Gonna Game Game Game Game Game
Analysis of chess openings says 1. e4 e5, while objectively strong at higher levels, underperforms at lower levels, provides other details. I think all of this is Doing It Wrong. If an opening was underperforming by 100+ points of Elo that would be one thing, but these differences are all small. Different openings teach you different things, help you develop in different ways and most of all play different types of games. I always played 1. e4 not because I thought it was an extra +1% to my expectation, but because I enjoy those games more.
Phil Nguyen provides Premodern deck pics for all his opponents. What a format. Also yeah, I am sad I won’t ever get to play with it but maybe it was time to ban Land Tax.
Sam Black comes for old school. Says Fireball should have been involved somewhere.
His write-up of his thought process is here. I am confident this would be better if we cut it down to 60, despite concerns over having enough Land Tax targets, there are only four semi-dead cards (Tormod’s Crypt, Millstone and Timetwister, and maybe Recall) that you want to avoid drawing too early and can’t safely cut given Divine Offering exists and you can’t run a cardless alternative damage source like Mishra’s Factory. I agree with Sam that there should be a Fireball somewhere, and Triskelion seems like definitely the wrong creature. My instinctive choice is to keep it old school and choose Serra Angel here.
If I did find extra space what would I add? I would need to know more about exatly what opponents were out there but my first priority would be access to a second Winter Orb and then a second Wrath of God or The Abyss, or I might think more about beating The Deck in particular.
Sam Black also shares his thoughts on the playing-cards-facedown-as-lands mechanic. He is against it, as it adds painful decision points and reduces variance.
Ben Seck is back with deck name (and deck pic!) thread part two. Original thread here.
Brian Kibler makes the case that Wizards needs to be pushing Standard play far harder than they are, which seems right to me. Otherwise, when I play constructed, I am not going to be giving Wizards money and things won’t change over time – others use Commander for this, I’d likely end up in Premodern.
Angel Hernandez, somehow only the second worst MLB umpire, decided to sue MLB over being passed over for crew chief and the world series, charging racial discrimination. Too much discrimination is definitely not his problem.
Not only game design.
A lot of this is an artifact of the player needing to maintain a very high win percentage, and often to avoiding even taking losses. If you are 50% to win a fight, after which you full heal, you cannot afford to be planning too much around 1% (or the more standard 1/16 or 1/20) crit chance in the wrong spot. If you need to win 10 or more fights in a row to finish the level, or a death costs you a lot of progress, not so much. If you’re playing on hardcore in Diablo, you have to assume the absolute worst case scenario, including multiple crits.
Whereas when you are retrying a boss fight over and over until you win, or you’re playing a roguelike on a level where you expect to lose, or you realize that playing it safe here is otherwise locally doomed, you do the opposite, and try to get lucky.
Now consider real life. In real life, dead is dead, you need to string together a 99%+ win rate on many fronts and the tail risks of almost every social interaction vastly exceed any potential benefits, with notably rare exceptions that get played very differently.
Thus it is vitally important to remove the possibility of such tail risks whenever possible, unless this is an impact that you want.
Introducing ESPN Bet, a sportsbook from ESPN. Oh no? Oh yeah! There is no world where this goes well, but there are plenty where it is hilarious.
Dota bans 90,000 smurf accounts, meaning alternative second accounts they were able to link back to an original, with plans to sanction the original accounts as well. Many smurfs are used to beat up on weaker players, to be abusive in chat or grief in various ways, or otherwise act in toxic fashion. Whereas they would prefer you pick on someone your own size and remain mindful of your reputation.
This seems good. In a game this skill testing, and it is massively skill testing, you need to ensure fair matchups and fair play via reputation to get a good experience. However, there is also a legitimate interest in experimenting or relaxing, without having to be ‘on’ all the time, as chess players know well. If your rating is always at risk, you can’t try anything new or different. You can say ‘well then play unranked’ but then conditions are presumably a hellscape, so you then have to assemble the entire lineup for both sides yourself. Alas, this remains an unsolved problem.
I Was Promised
FlyingSelf-Driving CarsSelf-driving cars got approved for San Francisco.
What is next? The driverless bus, of course.
Is this our future? A driverless bus, in a loop, with someone paid to sit there and not drive it, forever.
Sure sounds like it:
There is also lots of talk about how this totally is not a replacement for the bus system, only for the first or last mile problem.
It seems like everyone is determined to ensure that no inefficiencies are lost in the transition to this new technology.
The San Francisco Board of Supervisors has thus formed a new plan to try and stop autonomous vehicles, which is to deny all fleet vehicles access to charging stations.
Also California Senate Appropriations Committee unanimously approves Bill AB 316, requiring any vehicle over 10k pounds have a driver physically present at all times on public roads.
Joe Weisenthal speculates about who else might not take kindly to the trend.
I mean, that’s not a trade I would have made, but yes the beeping is super annoying and condescending as hell and I too despise it when cars and other things try to punish you into cooperating with their agendas. A fully self-driving car that might get hacked, or that the government might potentially take control over, seems terrifying on many levels. For a taxi it’s fine, for your own car you better be damn sure the override button works and they don’t have an override override of their own. Or you can say the opposite, and put safety or control first, but then yes you are going to have a problem.
On the liberal side, so far what I am seeing is not tech skepticism so much as fear about job losses, direct opposition to the idea that we might produce higher quality goods while using less labor, combined with safety illiteracy likely motivated by the jobs issue.
The other weird question I have is, why would you use a beep as your plan to ensure someone doesn’t kill the car battery, as opposed to it turning off on a timer unless you hit the ‘no I really mean it’ button?
Timothy Lee writes under the headline driverless cars may already be safer than human drivers, and makes the case that actually we do know, and they are, by going through every single crash history report for Waymo and Cruise in California.
Cruise’s history is less impressive, still solid, with three crashes where the Cruise car was clearly at fault that were more serious than any of Waymo’s mistakes, while noting Cruise has clocked more miles. He doesn’t quite say ‘self-driving cars are safer than human drivers’ but I will say it: Self-driving cars are safer than human drivers, sufficiently so that if we previously had self-driving cars no way would we let humans drive a car.
Having full video of every incident makes it tough to argue. The New York Times is undaunted, says self-driving car (briefly) delayed ambulance attending to… someone hit by a human driving a car. Yes, it turns out that cars on roads delay other vehicles sometimes. Can you imagine this standard being applied to human drivers? Or if a self-driving car had hit a driver, a human had briefly gotten in the way of the ambulance?
While I Cannot Condone This
The cancellation attempt on Richard Hanaina appears to have failed across the board. More evidence that woke has peaked.
Chrome now by default using your browser history to show you personalized ads.
You can also go here, I believe, which also lets you do more customization.
Except, good? I love personalized ads. I especially love that I can manually ask for more or less of various categories, including blocking anything I find unpleasant, distracting or likely to get me to buy something I’d regret. I do get that there are people who either want to see the same ads as everyone else, and those who think ads are evil or adversarial so you want them to be as irrelevant as possible. I think they are mostly wrong, or are pointing to a deeper problem they should address, although in some cases the solution will indeed involve blocking some types of ads.
What if Zoom had a button where if all participants press the button then the meeting instantly ends?
I would presume the correct strategy is the obvious one, which is that you press it if you want the meeting to end conditional on them also wanting the meeting to end. Unless of course you think you’d get in trouble if the meeting actually ended, they were only testing you by pressing it, whoops. There is that. Trust can be hard.
A potential variation is to use the 10-point scale of activity, where you both pick a number from 0-10 (0 by default) and if they add to 10 or more then you do the thing, and you choose such that the decision is made correctly. Then the game theory seems clear, you mostly enter the number that matches your desire to be there, but you only have to ‘spend the points’ on that if they try to end the meeting, at which point they know you have a higher number entered.
Is reason the slave of the passions, or does reason tame the passions, or do the passions make you bad at reason?
Gaven Tredoux thread emphasizing that the ‘fellow travelers’ in the West were not unknowingly duped by the USSR, they were active participants in propaganda and information suppression campaigns on a massive scale. There has been a complete lack of reconsideration or condemnation of such figures, which is important context when people reconsider and condemn other historical figures for things far milder.
Taleb proposes that books that endure are almost always poorly written, but address fundamental topics. Other books, often even by the same author, focus on other things and get discarded.
There is something to this. As always, there can only be one top priority. One’s mind can be focused first on the ideas, or focused first on the presentation. If a writer focuses too much on ‘getting good at writing’ they won’t be getting good at ideas. The two needs will often directly clash, and you’ll have to choose, and it will show.
So I think the dilemma holds true at the top. If you are ‘masterfully crafting’ your book, it is unlikely to be an important one. If your book is important, better to ensure you get the important things across than craft masterfully. However, I do not think this holds more generally. Books that are badly written tend to simply be bad, all good things and skills tend to positively correlate, when you know you have something worthwhile you put in more effort, as I do in my occasional ‘effort posts.’
Short video showing a social selling factory in Indonesia, with various booths for content creation. Seems highly efficient, love the ingenuity, seems like a nice classy place where people are having a blast. Give the people what they want, at least until the AI can give it to them cheaper. The one thing I can’t figure out is why they don’t need vastly better soundproofing.
The Lighter Side
PornHub threatens to sue local kabab restaurant Döner Haus to make them change their name, lest someone looking for hardcore porn via their ‘Doner Haus Porn Videos’ instead end up instead buying a sandwich. I am excited to try one of these two products.
An economics lesson.
Another economics lesson.
Patrick McKenzie explains to his wife how America tracks its citizens. Worth it.
The state of local politics.
They took our jobs.
It may not be the AI meme maker the world needs, but it is the one it deserves.
You say unhinged. I say plausible justification of a 50% workload reduction.
You say unhinged, I say they hate us for our freedoms, an ongoing immigrant saga.
Like it or not, we are so back.
Except, oh no, they are on to us.
I would also note that in general, if I am doing something like this where I likely don’t get much inherent value from the work, I’d much rather make $15k at 500 a pop than make $45k at 100 a pop. Team Daniel all the way.
A story of true resourcefulness (5 minutes).
A story of less resourcefulness.
The ultimate story of resourcefulness.