I would not be at all surprised if someone told me that the deadweight losses from signaling swallowed 10% of society’s productive capacity.
I was surprised to see such a low figure. I would put the number much, much higher.
I’d bet than, at a typical randomly selected company, they could achieve their objectives for close to an order of magnitude less cost if the employees actually cared about the objectives in an actually trying sort of way. I’m getting this 90% inefficiency by judging the difference in cost between companies with an objective it’s employees actually care about and companies with phony objectives the employees aren’t so emotionally invested in. (For example, SpaceX vs Boeing or Lockheed Martin.) I don’t think it’s quite that bad, but it seems to be the right order of magnitude.
I think most people are trying mostly to signal competence and hard work to their bosses. It’s far easier to look like you’re doing useful work than to actually do something useful. I can’t find the link at the moment, but I believe some survey showed that almost everyone would take a bet with 10:1 odds against winning if the payoff was $100 for every dollar bet. However, when asked whether they thought their boss would want them to take such a bet with company money, almost all said “no”. This hints that individuals within companies may pass up on risky or non-Ra-flavored opportunities.
In order to signal competence well in most jobs, you still have to be somewhat competent. Perhaps programming ability is hard to fake, but management consulting is easy to fake. If so, then in accordance with Goodhart’s law, a rational self-interested management consultant should spend almost all his or her time making spreadsheets as pretty as possible, learning all the buzzwords, keeping up with the latest management fads, and optimizing for whatever other subjective or objective metrics of success they may be judged by. If they do not, and waste time on actually achieving the company’s goals, they will quickly be out-competed by someone better at signaling.
But all that is just signaling losses within a company. There are entire industries devoted to signaling. If half of all GDP is for products that exist to signal, then we might estimate that 0.1*0.5=0.05=5% of all human activity is devoted to things besides signaling. Unfortunately, this is harder to estimate, since the big industries like mining and construction make products that serve a mix of signaling and non-signaling purposes. I’d be a little surprised if 90% of GDP was exchanged for signaling reasons, for a total signaling burden of 99%, but that doesn’t seem completely unreasonable to me. We can put an upper bound on it, though. Suppose the cost of 1 person living comfortably with 0 signaling is $15,000 a year or less, but this could be $1,500 a year without signaling bloat in industry. Per capita GDP in the US is $72,923, so that puts an upper bound of ~98% signaling. If you think the minimum amount you could live comfortable for is $20,000, and you think my 90% figure is inflated, then use 75% to get a cost of living of $5,000/year. That would suggest 93% of GDP goes to signaling.
(The efficient market hypothesis does make this argument less probable, but not impossible. It’s financially, socially, and legally difficult to monitor not just employees actions, to make sure they aren’t slacking, but also their thoughts, to make sure they are spending significant mental effort looking for better ways of doing things, rather than just blindly doing what’s mentally easiest on their System 1. Perhaps in the Age of EM we will have that level of monitoring, but at the moment it’s too expensive to go full 1984 on employees. The only good metric we have is the success of large projects, which suffers from coordination problems and free-rider problems. Some startups get around this by using the “change the world” line and convince employees to make company goals into personal goals. They then oust the old regime, and slowly turn into the next generation’s stagnant industry.)
This seems believable; but losses compared to what is the question.
It’s hard to maintain “actually trying” levels of effort to achieve objectives in a large organization, simply because it’s hard to communicate across large numbers of people what the objectives are.
However, large organizations are still economically efficient for some uses. An economy consisting only of five-person startups would, I’m pretty sure, be poorer than the one we live in today.
It seems that we trade off “degree of honest goal-alignment” with “returns to scale” to some extent.
The new research also shows that monopolists typically increase prices by using political machinery to limit the output of competing products—usually by blocking low-cost substitutes. By limiting supply of these competing products, the monopolist drives up demand for its own. Thus, in contrast to conventional theory, the monopolist actually produces more of its own product than it would in a competitive market, not less. But because production of the substitutes is restricted, total output falls.
He says that this is because, contrary to the standard view of monopolies as homogeneous and rational actors, they are actually composed of many subgroups vying for power. Sometimes this is within a single organization, or sometimes a similar effect can be created by large groups of independent practices banding together. Lawyers band together to uphold legislation keeping people from offering legal advice without passing the bar exam. Construction unions prevent the use of prefabricated parts.
I can’t find the link at the moment, but I believe some survey showed that almost everyone would take a bet with 10:1 odds against winning if the payoff was $100 for every dollar bet. However, when asked whether they thought their boss would want them to take such a bet with company money, almost all said “no”.
If you still find the source I would be interested.
Normally, I’m pretty good at remembering sources I get info from, or at least enough that I can find it again quickly. Not so much in this case. This was about halfway through a TED talk, but unfortunately TED doesn’t search their “interactive transcripts” when you use the search function on their page. A normal web search for the sorts of terms I remember doesn’t seam to be coming up with anything.
I scanned through all the TED talks in my browser history without much luck, but I have this vague notion that the speaker used the example to make a point about the importance of risk taking or something. But that doesn’t really narrow down the search space within TED much, so I can’t use it as a heuristic to screen search results.
Unless you want to scan through a couple hundred or thousand TED transcripts, or know of a way to search TED for keywords not in the titles, I’m ad a dead end. Sorry.
Yeah, some work is for reals, and other work is makework.
And this gives rise to the brilliant idea of getting rid of all of the makework. Cut out ALL the middlemen. Sounds great.
But you are playing with fire there.
Like, middlemen...are men. Useless people....are people. If you take away their meal ticket they will be annoyed. Annoyed humans are the most dangerous things that exist.
I was surprised to see such a low figure. I would put the number much, much higher.
I’d bet than, at a typical randomly selected company, they could achieve their objectives for close to an order of magnitude less cost if the employees actually cared about the objectives in an actually trying sort of way. I’m getting this 90% inefficiency by judging the difference in cost between companies with an objective it’s employees actually care about and companies with phony objectives the employees aren’t so emotionally invested in. (For example, SpaceX vs Boeing or Lockheed Martin.) I don’t think it’s quite that bad, but it seems to be the right order of magnitude.
I think most people are trying mostly to signal competence and hard work to their bosses. It’s far easier to look like you’re doing useful work than to actually do something useful. I can’t find the link at the moment, but I believe some survey showed that almost everyone would take a bet with 10:1 odds against winning if the payoff was $100 for every dollar bet. However, when asked whether they thought their boss would want them to take such a bet with company money, almost all said “no”. This hints that individuals within companies may pass up on risky or non-Ra-flavored opportunities.
In order to signal competence well in most jobs, you still have to be somewhat competent. Perhaps programming ability is hard to fake, but management consulting is easy to fake. If so, then in accordance with Goodhart’s law, a rational self-interested management consultant should spend almost all his or her time making spreadsheets as pretty as possible, learning all the buzzwords, keeping up with the latest management fads, and optimizing for whatever other subjective or objective metrics of success they may be judged by. If they do not, and waste time on actually achieving the company’s goals, they will quickly be out-competed by someone better at signaling.
But all that is just signaling losses within a company. There are entire industries devoted to signaling. If half of all GDP is for products that exist to signal, then we might estimate that 0.1*0.5=0.05=5% of all human activity is devoted to things besides signaling. Unfortunately, this is harder to estimate, since the big industries like mining and construction make products that serve a mix of signaling and non-signaling purposes. I’d be a little surprised if 90% of GDP was exchanged for signaling reasons, for a total signaling burden of 99%, but that doesn’t seem completely unreasonable to me. We can put an upper bound on it, though. Suppose the cost of 1 person living comfortably with 0 signaling is $15,000 a year or less, but this could be $1,500 a year without signaling bloat in industry. Per capita GDP in the US is $72,923, so that puts an upper bound of ~98% signaling. If you think the minimum amount you could live comfortable for is $20,000, and you think my 90% figure is inflated, then use 75% to get a cost of living of $5,000/year. That would suggest 93% of GDP goes to signaling.
(The efficient market hypothesis does make this argument less probable, but not impossible. It’s financially, socially, and legally difficult to monitor not just employees actions, to make sure they aren’t slacking, but also their thoughts, to make sure they are spending significant mental effort looking for better ways of doing things, rather than just blindly doing what’s mentally easiest on their System 1. Perhaps in the Age of EM we will have that level of monitoring, but at the moment it’s too expensive to go full 1984 on employees. The only good metric we have is the success of large projects, which suffers from coordination problems and free-rider problems. Some startups get around this by using the “change the world” line and convince employees to make company goals into personal goals. They then oust the old regime, and slowly turn into the next generation’s stagnant industry.)
This seems believable; but losses compared to what is the question.
It’s hard to maintain “actually trying” levels of effort to achieve objectives in a large organization, simply because it’s hard to communicate across large numbers of people what the objectives are.
However, large organizations are still economically efficient for some uses. An economy consisting only of five-person startups would, I’m pretty sure, be poorer than the one we live in today.
It seems that we trade off “degree of honest goal-alignment” with “returns to scale” to some extent.
Arguably, monopoly power artificially makes organizations larger than they would naturally be, which makes a higher share of the economy into signaling/makework than it optimally would be. See https://www.minneapolisfed.org/publications/the-region/the-costs-of-monopoly-a-new-view, and, like, all of Gabriel Kolko.
404: File Not Found. I think LW added the comma to the URL. This link should work though, for anyone interested:
https://www.minneapolisfed.org/publications/the-region/the-costs-of-monopoly-a-new-view
TL;DR:
He says that this is because, contrary to the standard view of monopolies as homogeneous and rational actors, they are actually composed of many subgroups vying for power. Sometimes this is within a single organization, or sometimes a similar effect can be created by large groups of independent practices banding together. Lawyers band together to uphold legislation keeping people from offering legal advice without passing the bar exam. Construction unions prevent the use of prefabricated parts.
If you still find the source I would be interested.
Normally, I’m pretty good at remembering sources I get info from, or at least enough that I can find it again quickly. Not so much in this case. This was about halfway through a TED talk, but unfortunately TED doesn’t search their “interactive transcripts” when you use the search function on their page. A normal web search for the sorts of terms I remember doesn’t seam to be coming up with anything.
I scanned through all the TED talks in my browser history without much luck, but I have this vague notion that the speaker used the example to make a point about the importance of risk taking or something. But that doesn’t really narrow down the search space within TED much, so I can’t use it as a heuristic to screen search results.
Unless you want to scan through a couple hundred or thousand TED transcripts, or know of a way to search TED for keywords not in the titles, I’m ad a dead end. Sorry.
I’m always torn when I see something like this.
Yeah, some work is for reals, and other work is makework.
And this gives rise to the brilliant idea of getting rid of all of the makework. Cut out ALL the middlemen. Sounds great.
But you are playing with fire there.
Like, middlemen...are men. Useless people....are people. If you take away their meal ticket they will be annoyed. Annoyed humans are the most dangerous things that exist.