A quick econ FAQ for AI/ML folks concerned about technological unemployment


by Eliezer Yudkowsky Mar 22 2017 updated Aug 18 2017

Yudkowsky's attempted description of standard economic concepts that he thinks are vital for talking about technological unemployment and related issues.

This is a FAQ aimed at a very rapid introduction to key standard economic concepts to professionals in AI and machine learning who have become concerned with the potential economic impacts of their work in the field.

It takes a strong intuitive understanding of "comparative advantage", "aggregate demand", and several other core concepts, to talk sensibly about a declining labor force participation rate and what ought to be done about it. There are things economists disagree about, and things that economists don't disagree about, and the latter set of concepts are indispensable for talking sensibly about technological unemployment. I won't belabor this point any further.

An extremely compressed summary of some of the concepts introduced here:

Many readers will also have ideas about:

There is not universal agreement on how large these problems are, nor whether they should be top priorities. But standard economics does not regard these ideas as naive, so I won't discuss them any further in this FAQ. My purpose here is to introduce relatively more technical concepts, that are more likely to be new to the reader, and that counterargue views widely regarded by professional economists as being misguided on technical grounds.

This document is currently incomplete, and the Table of Contents below shows what is currently covered. Warning: as of March 2017 this is a work in progress and has not yet been reviewed for accuracy by relevant specialists.


Comparative advantage, and the mercantilist fallacy

Ricardo's Law of Comparative Advantage says that parties (individuals, groups, countries) can both benefit from trading with each other, even when one party has an absolute advantage over the other in every factor of production.


If Alice wants an equal number of apples and oranges, she's better off producing 200 apples with all of her labor, and then trading 100 of the apples to Bob for 100 oranges.

Bob in turn is better off producing 100 oranges with 1 unit of labor, and trading them to Alice for 100 apples, than Bob would be if he used 2 units of labor to produce 100 apples directly.

Economists usually regard this idea as refuting some widespread intuitions about mercantilism, which says that countries are better off the more they sell to other countries, and worse off the more they buy from other countries.

Another possibly-not-intuitive consequence of comparative advantage is that, except for sticky prices (see below), the exchange rates on a country's currency have little long-run effect on its trade.


One day the European Central Bank decides to print enough new euros to halve the value of the currency. Then:

Ricardo's Law thus suggests that we can view the US as shipping 100 air conditioners to Germany in exchange for 10 cars, with the currency exchange rates being something of an epiphenomenon. (In practice this is modified by "sticky prices" in that some people already own euros or dollars, or bonds denominated in euros or dollars, and there are existing sales contracts that fix prices for a term. But effects like that tend to wash out over the longer run.)

The concept of comparative advantage is important for discussing technological unemployment because it conveys the general idea that when we draw a line around a group or region, like "{Alice, Bob, and Carol}" or "Wisconsin", this collective entity is not harmed by being able to trade with someone else who is more productive than them. (To say nothing of getting their own advanced technology.)

This isn't to say that technological advances can never hurt Wisconsin's economy. Maybe Wisconsin is producing cheese and trading it to Ohio in exchange for Ohio's coal. One day, Michigan becomes much more efficient than Wisconsin at producing cheese, causing Ohio to start trading with Michigan instead. This can hurt Wisconsin's economy, quite a lot in fact.

The conclusion from Ricardo's Law is rather that Wisconsin can't be harmed by the fact of Wisconsin itself being allowed to trade with Michigan. What hurts Wisconsin in the above scenario is that Ohio is allowed to trade with Michigan. Wisconsin can only make itself even worse off by choosing to cut off its own trade with Michigan.

If Wisconsin closes its borders and tries to get along in its own little world that doesn't include those awful new cheese-making machines, Ricardo's Law says that Wisconsin as a whole should become worse-off thereby. If this stops being true, then some unusual phenomenon must be at work, one that violates the broad axioms under which Ricardo's Law is a theorem.

An intuition I'll be trying to pump throughout this FAQ is that the state of affairs over most of human history, in which technological automation made most people better off without causing them to be permanently unemployed, is a very normal state of affairs from the standpoint of economic theory. If we're currently looking at lasting unemployment caused by automation, this is a surprising state of affairs.

And: Even if something has changed and something weird is going on, the reason for it won't be that Ricardo's Law suddenly stopped being a theorem. Whatever the truth turns out to be, Ricardo's Law will still be a relatively better lens than mercantilism through which to understand it.

Complementary inputs, and the lump of labour fallacy

Broadly speaking, the "lump of labour fallacy" is reasoning as if there's a limited amount of work to be done in the world, and a limited number of jobs to go around, and every time one of those jobs is automated away, one more person ends up unemployed.

Suppose that:

So then the critical question becomes: if we invent a new kind of sausage-making machine that doubles productivity and makes it possible to produce 2 sausages using 2 units of labor, do we…

Ending up in the second equilibrium, maybe not right away, is widely regarded as the core idea behind how we could start with a world where 98% of the population were farmers, improving agricultural productivity by a factor of 100 between then and now, and end up with a world in which 3% of the people are farmers and the other 95% of the population are not unemployed.

One way of looking at this is that, over the course of human history so far, everything has behaved like it is complementary. When agricultural productivity skyrocketed, a bunch of people went into making shoes and clothes and tables and houses. The end result was that more people had food, more people had more nice clothes and furniture, and not that most of the population ended up unemployed. The "food" input to a human became cheaper, and labor went into making more of the "clothes" and "furniture" inputs to humans. Then again later, when manufacturing productivity skyrocketed, more people went into services.

Now it could be that this phenomenon is now breaking down for any number of possible reasons. But among those reasons, it seems relatively implausible that we have reached the end of demand and that all the people now have enough of every kind of good and service that they want. There are way too many poor and unhappy people still in the world for that to be true!

Okay, so we haven't reached the end of human wants. But maybe some people in the modern world can no longer produce anything that other people want…?

I'd like to encourage you to regard this idea with some surprise. Especially the notion that this "unnecessariat" exists today, before we can reliably build a robot that puts away the dishes from a dishwasher without breaking them. AI is making progress, but we haven't exactly reached par-human perception and motor skills, yet. Are there really people in the world who can do nothing that anybody else with money wants?

Heck, why can't those people trade with each other? Why can't people in this supposed unnecessariat at least help each other, and get their own economy going, even if nobody else wants to trade with them? And doesn't Ricardo's Law say that this state of affairs ought to provide a floor on the group's minimum economic activity, since in theory the group can only benefit further by trading with the outside world…?

Why isn't it an economic theorem that there's never any unemployment?

I would encourage you to stare at this argument and try to feel surprised that unemployment exists. It's not that theorems get to overrule reality; if a valid theorem fails to describe reality, that just says one of the axioms must be empirically false. But we do know that more must be going on in the existence of unemployment, than there being a limited lump of jobs and not enough jobs for all the people.

And so on. There are more than enough possible hypotheses to explain why unemployment can possibly exist. Still, it's worth noting that there was less lasting unemployment in, say, 1850. What changed between then and now… could be all the regulations saying you can't just go build somebody a house. Or it could be increased inequality. Or it could be housing prices preventing people from moving to California where there are jobs.

But it's relatively much stranger to postulate that the only problem is that we're running out of jobs in the global lump of labor as AI advances eat away at the lump.

It could indeed be that existing AI, and to a much larger degree, ordinary computer programming, has gone a long way towards devaluing many jobs that can be done without a college degree.

Labor force participation is in fact dropping under current conditions, whatever those are. It's possible and maybe probable that, all else being equal, AI conquering more perceptual and motor tasks will make more people permanently unemployed.

It's reasonable for people in AI to feel concerned about this and want to make things better.

But it's unlikely that the issue is the end of human want, or the running-out of the lump of labor. The problem may not be that some people have intrinsically nothing left to contribute to any remaining human want; but rather, that the modern world has put obstacles in the way of people being allowed to contribute.

Regardless: However you look at the situation, and whatever solution you propose, please don't talk about AI or trade or automation "destroying jobs." This will cause any economists listening to scream in silent horror that they cannot give voice.

The broken window fallacy, and aggregate demand

Surprisingly, recessions exist

In theory, when we double the productivity of sausage-makers, we should get 15 sausages-in-buns instead of 10. We should not get 10 unemployed sausage-makers; or, if we temporarily get 10 unemployed sausage-makers, that state of affairs shouldn't last longer than it takes sausage-makers to become willing to look for work at the booming bun factories with HELP WANTED signs plastered all over their windows.

And yet there seem to be these occasions, like the "Great Depression" or the "Great Recession", when there are silent factories and people standing idle, or the modern equivalent. The unemployed sausage-makers are willing to look for new jobs, but there aren't tons of HELP WANTED signs to let them get started on retraining. The bun factories are actually making fewer buns, because now the unemployed sausage-makers aren't buying sausages-in-a-bun anymore.

One will observe, however, that these historical occasions seem to come on suddenly, and not due to sudden brilliant inventions causing a huge jump in productivity. They happen at around the same time that there's trouble in the financial system--banks blowing up.

Furthermore, the number of idle factories and amount of idle labor seems to be far in excess of what could reasonably be accounted for by sausage-makers needing to change jobs. And even people with what might seem like very useful skills, can't find anywhere to put them to use. None of the previous reasons we've considered, for why unemployment could possibly exist, seem to apply to how that much unemployment could exist in the Great Depression. And then a decade later all those people had jobs again, too; so it wasn't something intrinsic to the people or the jobs that caused them to be unemployed for so long.

That is weird and astonishing! If your brain doesn't currently think that is weird and astonishing, please try at least briefly to get yourself into the state of mind where it is.

The broken window fallacy

As my entry point into explaining this astonishing paradox, I'm going to take an unusual angle and start with the broken window fallacy.

The broken window fallacy was originally pointed out in 1850 by Frederic Bastiat, in a now-famous essay, "That Which Is Seen, And That Which Is Not Seen" Bastiat begins with the parable of a child who has accidentally thrown a rock through somebody's window:

Have you ever witnessed the anger of the good shopkeeper, when his careless son happened to break a square of glass? If you have been present at such a scene, you will most assuredly bear witness to the fact, that every one of the spectators, were there even thirty of them, by common consent apparently, offered the unfortunate owner this invariable consolation — "It is an ill wind that blows nobody good. Everybody must live, and what would become of the glaziers if panes of glass were never broken?"

And doesn't the glazier, in turn, spend money at the baker, and the shoemaker, and the baker and shoemaker spend money at the mill and tannery? Doesn't all of society end up benefiting from this broken window?

Now, this form of condolence contains an entire theory, which it will be well to show up in this simple case, seeing that it is precisely the same as that which, unhappily, regulates the greater part of our economical institutions.

Suppose it cost six francs to repair the damage, and you say that the accident brings six francs to the glazier's trade — that it encourages that trade to the amount of six francs — I grant it; I have not a word to say against it; you reason justly. The glazier comes, performs his task, receives his six francs, rubs his hands, and, in his heart, blesses the careless child. All this is that which is seen.

But if, on the other hand, you come to the conclusion, as is too often the case, that it is a good thing to break windows, that it causes money to circulate, and that the encouragement of industry in general will be the result of it, you will oblige me to call out, "Stop there! your theory is confined to that which is seen; it takes no account of that which is not seen."

It is not seen that as our shopkeeper has spent six francs upon one thing, he cannot spend them upon another. It is not seen that if he had not had a window to replace, he would, perhaps, have replaced his old shoes, or added another book to his library. In short, he would have employed his six francs in some way, which this accident has prevented.

Bastiat's point was widely accepted as persuasive; or at least, it was accepted by economists.

Let us nonetheless notice that Bastiat's original essay is worded in a bit of an odd way. Why focus on the fact that when the shopkeeper has spent six francs in one place, he therefore can't spend six francs somewhere else? If the limiting factor is just money, couldn't we make all of society better off by adding more money?

You would ordinarily expect a healthy economy to be limited by its resources, by the available sand and heat for making glass. If the glazier replaces one window, he isn't replacing another. Or if some other seller provides the glazier with more sand and coal, to make more glass than he otherwise would've, then that's coal which isn't going to the smithy.

Of course what Bastiat really meant is that sending resources to repair a broken window calls away those resources from somewhere else. This is symbolized by the six francs being spent in one place rather than another.

That's how a market economy works, after all: when goods flow in one direction, money flows in the other direction. Six francs flow from the shopkeeper to the glazier; a glass window travels in the opposite direction.

So Bastiat talks about a change in a flow of francs, as a shorthand for talking about what really matters, a corresponding change in the opposite flow of goods.

Now ask: What if the economy is, somehow, Greatly Depressed? What if the glazier was standing idle until the window was broken, and isn't being called away from any other job? What if there's a ton of coal in an empty lot, waiting for someone to purchase it, and the reason nobody is purchasing it is that nobody seems to have any money?

Well, in this case, it still doesn't help to go around breaking windows. And now Bastiat's original wording happens to precisely describe the resulting problem: the shopkeeper will spend six francs on the broken window, and therefore not spend six francs on something else.

The only way we can imagine that breaking a window will help this economy… is if the shopkeeper has a buried chest of silver; and when somebody throws a rock through his window, the shopkeeper spends some of this buried silver; and then the shopkeeper doesn't try later to top off this chest of buried silver again. Only in that hypothetical is there actually an additional six francs added to the economy, flowing to an otherwise idle glazier, who trades with a baker that wouldn't bake otherwise, who buys from a farmer who can afford to buy more fertilizer and grow more crops than before.

Would you agree with that statement -- with the paragraph as written above? Standard economics says it is correct.

But this suggests an astonishing implication. It seems to say that we could just print more money and make the shopkeeper's town better off. Printing new money is pretty much the same as digging up money from a buried chest and not replacing it later… right?

More money, fewer problems?

"Now hold on," says the alert computer scientist, sitting bolt upright. "I know that society conditions us to think of little rectangular pieces of colored paper as super-valuable. But dollar bills or euro bills or whatever are just symbols for actual goods and services that are actually valuable. Thinking you can create more goods and services by creating more money is like counting your fingers, getting an answer of 10, erasing the 10 and writing down 11, and thinking you'll end up with 11 fingers. Money isn't wealth, it's a claim on wealth, and if you counterfeit $100 in your bank account, you're just stealing wealth that someone else won't get. Any theory that says you can create more wealth by creating more money, no matter how complicated the argument, will in the end turn out to be missing some row of the spreadsheet that makes the totals add up to zero."

And in a healthy economy, this would be correct. But remember, those idle factories are not supposed to exist in the first place. Is it so impossible that a weird condition that shouldn't exist, could be fixed by a weird tactic that shouldn't work?

"So what you're saying," says the skeptical computer scientist, "is that the two impossibilities cancel out. Like the story about the physics student who derives $~$E = -mc^2,$~$ and who gets told to try next time to make an even number of mistakes."

Well… yes, as it turns out. There's even an elegant explanation, which we'll get to, for why those two particular impossibilities are symmetrical and cancel each other out.

First, let's consider a more ordinary exception to the rule that creating money can't help--the reason that money exists in the first place.

Suppose that Alice, Bob, and Carol are as usual stuck on a deserted island. Alice is growing apples, Bob grows bananas, and Carol grows cucumbers. One day Alice wants a banana, Bob wants a cucumber, and Carol wants an apple; and they all go hungry because Alice doesn't have any cucumbers to trade to Bob for a banana. Also they don't have computers on their deserted island, so they can't do anything clever like write software to detect cycles in people's desired goods.

In this case the standard solution for this lack of computing power is… (drumroll) …money!

Creating symbolic money to add to this island, where there was no money before, can indeed the three people on the island materially better off. Inventing money causes more trades to occur than previously; trading can make people materially better off.

When an economy has idle factories and standing workers, it's in a state where more trades could occur, but aren't occurring.

So… adding more money to that economy, to animate more trades, isn't far from what happens when we add money to an island that doesn't have money?

Now I have, indeed, pulled some wool over your eyes here. The macroeconomic theory is waaaaay more complicated than this, in much the same way modern neural nets are a tad more complicated than just the idea of gradient descent.

To give an idea of some of the complications we are blatantly skipping over: classical economics says that destroying money should not in fact slow down an economy composed of rational actors. In a rational world, if somebody sets fire to half of the money supply, everyone just halves their prices and life goes on.

The reason our world is not like that ideal world where the prices just go down, is partially that there are outstanding loans denominated in the current currency, and existing contracts denominated in the current currency. But most of the problem--according to the particular economic school I happen to subscribe to--is that people who set prices are reluctant to lower them, in a way that an economy full of perfectly rational actors wouldn't be.

And this is the kind of statement that causes fistfights between economists.

But there is widespread agreement that, for whatever reason, we don't live in the classical world of perfectly rational actors.

And that is--on the standard theory--why fractional reserve banks blowing up and destroying money and making fewer loans, cause there to be a bunch of unemployed people and fewer jobs; and nobody being able to afford to buy things because none of their own customers can afford to buy from them. In a modern economy, for every flow of goods, there's money flowing the other way. So when you destroy money or slow down its flow, this can reduce the actual number of trades.

Which means that banks blowing up and thereby reducing the flow of money, can reduce the amount of flowing goods instead of prices dropping. We are pretty sure from observation that this does in fact happen.

After we're done clipping gradients and initializing orthogonal weights and normalizing batches, in the end deep neural nets are still sliding down a loss gradient. Similarly, it really looks in theory and in historical observation like an economy with idle resources, especially unemployed workers, can be made materially wealthier by creating more money, which causes more trades to occur and brings the economy closer to operating at full capacity.

Historical observation also seems to bear this part out, although the near-impossibility of running real controlled experiments makes it hard to be sure. The most recent case you might have heard about was in 2013 in the United States, where the "sequester" was going to produce a sharp cut in government spending just as the US was starting to recover a little bit, not completely, from the Great Recession. Some people predicted the economy would tank again; and other people said it would be fine so long as the Federal Reserve created a correspondingly large amount of money, aka monetary offset. And the sequester happened, and the Federal Reserve did in fact print a bunch of money, and pretty much nothing happened to the economy. Amazing proof, right?

Okay, so that's not a large number of bits of evidence by your standards, you spoiled brats who get to test neural nets on 127GB of training data. But it was in fact an experimental test of two different predictions, and the economists who thought in terms of flowing money were right; and in the end, that's all we have in this life.

A general glut?

John Maynard Keynes -- I know some of you are shuddering at the name, but he was a respected economist of his day and he came up with a clever illustration, so please bear with me -- John Maynard Keynes once came up with an illustration of what could possibly cause anything as weird as a recession, and how this could possibly be fixed by anything as weird as printing money.

Suppose that on an island, Alice and Bob and Carol are growing apples and oranges and pomegranates. Furthermore, as in the Great Depression, it is not yet the case that everyone has enough to eat.

In this situation, could there be an excess supply or "glut" of apples?

Well, if people are still willing to eat apples, then what would "a glut of apples" even mean?

We answer: Suppose Alice and Bob, who both grow apples, have such a large apple harvest that people on the island feel kind of saturated on apples, and now prefer to eat other, scarcer fruits if available. In this case we'd see the relative trading value of apples dropping, compared to other fruits. Maybe 1 apple used to be worth 1 orange, and now somebody would only trade 1 orange for 2 apples. This would signal Alice and Bob to shift more of their effort to growing oranges and pomegranates in the future.

We could perhaps call this a "glut" of apples, meaning a relative excess of supply of apples, versus the supply/demand balance of other fruits, compared to previously.

Observe that to say that apples are now cheaper in terms of oranges, is equally to say that oranges have become more expensive in terms of apples. Then is it possible to have a glut of every kind of fruit at once?

How (asked Keynes) could that possibly be? So long as Alice and Bob and Carol aren't stuffed full, so long as they would still prefer to eat more fruit, how could that possibly be? You can't have each fruit being ultra-cheap as denominated in all the other fruits.

Then what are we to think if we see Alice and Bob and Carol collectively cutting back on growing apples and oranges and pomegranates?

How is it possible for a whole economy to be suffering from a problem of "excess supply," a situation where factories in general are going idle, without there being a corresponding surge of demand somewhere else?

The answer (said Keynes) is that the only way this "general glut" can happen, is if there is some additional, invisible good that people are pursuing more than apples and oranges and pomegranates. What we see is the value of all three fruits falling relative to the value of this invisible good, so that people produce less of all three of them. And the way that this can happen in real life (said Keynes) is if Alice and Bob and Carol have invented a fourth, invisible good called money.

This may sound weird and abstract, so consider this even simpler real-life example drawn from the overused case of the Capitol Hill Babysitting Co-op, a baby-sitting club which created a currency that parents could trade among themselves to pay for babysitting. The parents in this system tried to keep a reserve of co-op scrip so they could be sure of getting babysitting on demand. In fact, it turned out, people wanted to keep more scrip than the co-op made available in total. Soon there was less and less actual babysitting going on as a result, and the co-op nearly died. Like a glut of apples that is symmetrically a boom in oranges, there was a glut of babysitting and a booming demand for babysitting tokens. Only the babysitting tokens by themselves, sitting around motionless, weren't really helping anything.

If you add in more goods, like oranges and apples and pomegranites, the overall situation is still the same: when the economy has spare productive capacity and yet everything seems to be slowing down at once, we can see the required symmetric boom as taking place in the demand for that final good of currency… which is quite harmful to the real economy, because currency doesn't do anything.

It's the weird nature of currency as a worthless symbolic good, that allows recessions and Depressions to happen in the first place. Which is why the spreadsheet can in fact balance, and the two symmetrical impossibilities can cancel out, when we try to fix a recession by creating more money!

Aggregate demand and monetary stimulus

When economists talk about this sort of thing, they often use the phrase "aggregate demand".

Actual human wants are, if not infinite, then certainly far higher than the total amount of stuff produced by the world economy. We don't all have mansions, at least not yet.

So what "aggregate demand" really means is "aggregate purchasing power", which in turn means the total amount of flowing money trying to buy all the goods.

The standard economic view can then be stated thus: when the purely symbolic financial system blows up, and yet in the real world people lose their jobs and factories stop operating, what's happening is an "aggregate demand deficit." There isn't enough purchasing power to buy all the stuff the economy could supply at full capacity.

The obvious reason this happens is that exploding banks (or nowadays, banks that get bailed out, but are less enthusiastic about making more loans afterwards) reduce the total "money supply." There's less total symbolic money to flow through the trades.

Another reason aggregate demand decreases, is that people want to hold on to more money during a depression or a recession. They have an increased preference for larger numbers in their bank accounts, and are more reluctant to spend money. This decreases monetary velocity.

When all this is starting to go wrong inside a country, conventional economics says that the central bank ought to immediately swing into action and create more money, aka "monetary stimulus." Or rather, the central bank is supposed to create money and use it to buy government bonds or something, and then hold onto the bonds. Later on, when monetary velocity picks up, the central bank should sell the bonds and destroy the money it previously created, to avoid creating too much inflation.

If the central bank thinks it can't create enough money, then a widely advocated policy says that the country's government should instead sell treasury bonds, in exchange for money, and spend that money on… pretty much anything. This is not in fact a null action in monetary terms--we are not just removing taxes and spending them elsewhere, or so most economists think. The people who buy bonds in exchange for money have bonds afterwards instead of money; and this can also satisfy their desire to hold assets. Then the money goes out the other end of the government and into circulation, where it wasn't circulating before. This is "fiscal stimulus."

Whether it is ever a good idea to do "fiscal stimulus" is another one of those questions that cause economists to get into fistfights. I personally happen to side with the school that goes around saying, "What do you mean, the central bank can't create enough money to stabilize the flow of currency and trade? Did you run out of ones and zeroes? Why are you asking the government to increase the national debt over this? Just do your darned job!"

I expect that this business of talking about how central banks are supposed to stabilize the national flow of money, is causing half of you to raise skeptical eyebrows and ask if maybe the whole thing would somehow stabilize itself if anyone could issue their own currency instead of it being a government monopoly.

And the other half of you are trying to figure out some way to solve it using a blockchain.

But right now, nearly all countries have central banks. So long as that is in fact the way things are, conventional economic theory says central banks should try to stabilize the flow of money, and hence wages, and hence employment, by constantly tweaking the amount of money in circulation.

It's not agreed among economists which countries today might be suffering from too little aggregate demand, and working under capacity. The economists in my preferred school suspect that it is presently happening inside the European Union due to the European Central Bank being run by lunatics. Most economists think the United States is not currently bounded by an aggregate demand deficit, or running far under capacity.

I myself feel a bit unsure about that. Sometimes it really seems like we could all be much happier if we all just simultaneously decided to be 10% richer. Or that, say, Arizona, would benefit a lot from having its own currency to use on the side, until everyone there was working for everyone else. But I haven't paid enough attention to the specifics here, and you probably shouldn't listen to me.

So: If you're worried about technological unemployment due to AI advances, one of the obvious things that we should do along with any other measures we take -- that we really really need to do, all this abstract stuff has an enormous in-practice impact on the real economy -- is make sure that people are rich enough to buy things.

It is possible for an economy to end up in a state where most people feel poor, and can't afford to buy things from each other, and go around being sad and out of work… such that you could in fact cause everyone to perk up and trade with each other and be happy again, just by declaring that everyone has more money. (Basic income would just move money around; the idea here is that you can make things better if everyone simultaneously has more money.) If everyone thinks they're too poor to afford to hire one another, you can sometimes fix the problem by just having everyone be rich instead of poor. If, after great advances in automation, you saw lots of people standing around doing nothing, that would be the first thing we ought to try.

I'm going to temporarily stop going on about this because there's other economic concepts I want to run through and I don't want to risk boring you about this one topic. We'll pick up this thread again in a later section about "NGDP level targeting."

For now, I just want to observe that, compared to problems like reforming labor markets, it can be a lot simpler to just print more ones and zeroes at the central bank. And central banks are not entirely unwilling to hear about it. It's arguably the best choice for what's worth spending political capital to fix first.

Gently sloping supply curves, and skyrocketing prices

By now you might be starting to imagine a rosy post-automation scenario which seems in theory like it shouldn't be that hard to achieve.

You could look further ahead than I think we're actually going to get in practice -- not because I don't think AI can get that far, quite the opposite really, but still -- and imagine a world where robots are far more common, and human beings… are still trading with each other.

A world where, if Alice knows how to grow food, and Bob knows how to build houses, then they're not both going around cold and hungry because agriculture and house-building can be automated.

A world where we haven't wantonly defined ourselves into poverty; a word where robots exist, but we've decided to be imagine enough pretend symbolic money into existence for us to afford them.

A world where, to be honest, some of the dignity of work has been lost. Where the most common new jobs aren't as awesome as forging trucks out of molten steel with your bare hands. But people still want things, and other people do those things and get paid for them.

A world where people have as at least as much self-respect, where the economy is at least as vibrant, as it was in say 1960. Because logically, any given group could always draw a line around themselves and go back to 1960s technology. And if that group then removes this imaginary line, starts trading with others, and adopts more technology, then life should only get better for them.

Because of Ricardo's Law of Comparative Advantage.


So… there are a number of obvious difficulties that could completely blow up this rosy scenario.

One such class of difficulties is if there's any good X that people really need, and that stays extremely expensive in the face of automation--if it takes months and months of babysitting to afford one unit of X.

Like, say, housing, health care, or college.

You can find various graphs showing that all of the increased productivity over the last 20 years has been sucked up by increasing housing prices. Or that all of the missing wages, as productivity rises and wages stay flat, can be accounted for by the cost to employers of health insurance. I can't recall seeing a graph like that for college costs and student debt, but it sure ain't cheap. If all of those graphs are true simultaneously (and they are), you can see why people might feel increasingly impoverished, even if their nominal wages in dollars are theoretically flat.

And if those costs, or any other costs, went on rising while a hypothetical tidal wave of automation was crashing down, that could smash any utopian vision of people living peacefully on less labor in a less costly world.

For a brief but excellent overview of the problem, I would recommend Slate Star's "Considerations on Cost Disease". It turns out, for example, that the United States spends around \$4000 per citizen on Medicaid and Medicare and insurance subsidies. \$4000 is roughly what other developed countries pay for universal coverage. It's not that the United States is unwilling to pay for universal coverage, but that somehow healthcare costs far more in the United States. (Without any visibly improved medical outcomes, of course.)

There are various aspects of all this that economists will still get into fistfights over, so I can't just hand you a standard analysis and solution. But there's at least one important standard economic lens through which to view the problem.

Supply and demand are always equal

If we look at all the loaves of bread sold in a city on Monday, then, on that Monday, the number of loaves of bread sold, and the number of loaves bought, are equal to one another. Every time a loaf of bread is sold, there's one person selling the loaf and one person buying a loaf.

Supply and demand are always equal.

Imagining supply not equaling demand is like imagining a collection of nonoverlapping line segments with an odd number of endpoints.

Or at least, that's how economists define the words "supply" and "demand".

So what use is such a merely tautologous equation?

Defining the terms that way lets economists talk about supply curves and demand curves as a function of prices. Or to be snootier, supply functions and demand functions. Since supply and demand are always equal, the point where the two curves meet tells us the price level.

"Hold on," you say. "If supply equals demand by definition, then how would we talk about the old days of the Soviet Union, where there were 1000 people who wanted toilet paper and only 400 rolls of toilet paper? And it was illegal for the store to raise the set price of toilet paper? In this case, didn't demand just directly exceed supply?"

What was seen in Soviet supermarkes were long lines of people waiting outside before the store opened, trying to get to the toilet paper before it's all gone. In this case, the time spent in line is by definition part of the 'price' of a roll of toilet paper. No matter how the demand curve slopes--no matter how much people want toilet paper, and how reluctant they are to give up--the time required to stand in line will increase until 600 people give up and go home. And then the monetary price of the toilet paper, plus the time required to stand in line, is the "price" at which the demand function equilibrates with the supply function.

This system wasn't good for the Soviet Union because the time spent in line was burned, destroyed, in a way that produced no additional goods. When the price is allowed to go up under capitalism, the buyers may indeed spend more money; but just passing that money around doesn't destroy wealth the way that standing in line destroys time. (The flip side is that everyone ultimately has an equal supply of time and not everyone has an equal supply of money. Poorer people can sometimes benefit relative to non-poor people when part of a good is repriced in time rather than money. Although of course sufficiently rich people will just hire someone else to stand in line.)

If you set up a cart selling \$20 bills for \$1 each, a line will form in front of the cart and extend until it's long enough to burn \$18.99 worth of time, as priced by people who could otherwise earn the least amount of money per hour. You can't actually sell \$20 bills, to anyone who wants them, at a price of $1, and have that be the real entire price.

A similar dynamic plays out every year when the organizers of Burning Man once again try to defy the laws of mathematics as well as capitalism in order to sell 70,000 tickets at a fixed price of \$500, into a market of way more than 70,000 people who are willing to pay significantly more than \$500. They also try really hard to outlaw resale at higher prices; if Burning Man finds out about the resale, your tickets invalidated, and your car will be turned around at the last minute and told to go home. The resulting scramble… well, I've heard it suggested that, in this case, the real "price" of the ticket is pulling on your social connections in order to obtain Burning Man tickets for a properly nominal price of \$500, via trading favors with people who were savvy enough to sign up for the sales lottery. Thus selecting socially savvy people to attend Burning Man. I've never gone there.

When subsidy is futile and will be assimilated

Imagine a city under siege, where there are only 3,000 loaves of bread to be had, and each loaf is selling for 2 golden crowns. Can the governor of the city ease the plight of the people by ordering the treasury to pay 1 golden crown for loaf? The people will still be hungry -- there's still only so much bread to go around -- but perhaps their financial sorrows can be eased?

No, in fact. The price of the 3,000 loaves of bread rises until all but 3,000 people drop out of the market for purchasing it. If there are 3,000 people willing to pay 2 golden crowns for a loaf of bread, and the governor steps in and says the city will pay 1 golden crown per loaf of the price, then there will be 3,000 people willing to pay 2 golden crowns plus 1 golden crown subsidy. The price of each loaf rises to 3 crowns, and the 3,000 people buying it are 2 golden crowns out of pocket, the same as before.

In economitalk, you'd say that the supply was inelastic: this supply curve is entirely horizontal with respect to price. No matter how much more people pay (the x axis), the supply of bread stays the same (the y axis).

The demand curve does slope, in this example; it slopes downward as usual with increasing price, until the demand curve crosses the flat supply curve at a price point of 2 golden crowns.

If the governor then adds a subsidy of 1 golden crown to each purchase of bread, we are in effect shifting the demand curve to the right along the x-axis - the demand for bread at 2 golden crowns, is now the demand level we would have previously seen at a price of 1 golden crown. We have had a change of variables: Demand(x') = Demand(x - 1 crown). Since the supply curve is flat, this just means that the two curves meet at a price' which is 1 golden crown greater than the old price.

In other words: It doesn't help any buyers to try to subsidize a good in inflexible supply. The only way subsidies can possibly help buyers at all, is if increasing the price of the good causes more of the good to exist.

And by this I don't just mean that we check whether the supply of a good is allowed to increase, and if it can increase, then subsidies are allowed to help people. When you hand out subsidies evenhandedly to anyone who wants to buy X, the only mechanical means by which the people receiving the checks have any more money at all is through the medium of increasing the supply of X. Everything else is just a change of variables and shifting the demand curve to the right.

The idea that the person getting a check for \$120 is better off because they now have \$120 is entirely an illusion. The only mechanical means of transmission by which this person has more money in the bank, at the end of the day, is whatever extent the supply of X goes up; and their bank accounts will end up at the same level as a strict function of the supply of X, irrespective of the nominal amounts listed on the checks and whether the subsidy checks are taxed. To whatever extent the supply of X is not increasing, the act of placing money into people's hands has exactly as much effect on their bank accounts as praying to a golden statue of a twenty-dollar bill (stipulating arguendo that we live in a universe where praying to golden statues of things doesn't work).

And yet you will observe that in all public political discourse that makes it onto TV, all the sober talking heads in business suits are talking as if by subsidizing people with \$120 checks we are causing their bank accounts to go up by \$120, rather than talking about how many new universities or doctors or houses the \$120 checks will cause to exist.

This is genuinely crazy. This is not a Republican-economist point or a Democratic-economist point. I know of no framework of economics in which cutting everyone in the market the same subsidy check makes them have more money at the end, except insofar as supply happens to increase and their retained money is a strict function of supply. That the sober talking heads on TV are talking otherwise is mass civilizational insanity. It's not like the conversation about global warming where at least in principle we could be wrong about the empirical effect of increasing carbon dioxide on future global temperatures. It's not even like evolutionary biology, where at least the enormous mountain of empirical evidence is complicated enough that you might need to spend a few days reading to understand. The public conversation about subsidies is patently illogical. It is as if the sober talking heads on TV in their business suits were having deep conversations about whether to pray to a golden statue of a \$20 bill or a \$100 bill. There is no mechanical effect on bank accounts of universal subsidy checks that is not entirely mediated by, and quantitatively the sole function of, the final supply level of goods.

Skyrocketing prices

A price can only skyrocket when a gently sloping demand curve meets a gently sloping supply curve.

If the demand curve slopes sharply downward, this describes a state of affairs where, as the price of the good increases, lots of people rapidly become unwilling to pay and drop out of the market before the price has moved very far.

If the supply curve slopes sharply upward, the price can't increase very far. As the price goes up, the market is rapidly flooded with more sellers seeking to produce the good.

But if the city of San Francisco refuses to build more than a handful of apartments, and a large number of people in search of jobs really want to live in San Francisco and do not want to be driven out even as prices go up… then prices go way up. Until, however reluctantly, the least wealthy people are driven out.

"I already knew that," you say. Okay, but now consider health care and college.

In the United States there are, if I recall correctly, only 350 people allowed to become orthodontists every year. That is how many residents the orthodontic schools have decided to accept. Everyone else gets turned away.

For as long as that deeply entrenched system holds, it can accomplish literally nothing to try to subsidize orthodontics. You cannot cause more poor children to have braces by having the government offer to pay part of the costs of orthodontia. 350 orthodontists per year can only put braces on a limited number of children. No amount of shuffling money around or tweaking insurance regulations can allow more children to have braces than that.

Once upon a time after World War II, the US government passed a GI bill subsidizing college for military veterans. And in actual practical reality… a lot more people ended up going to college! So there must have been more colleges springing up, or existing colleges must have expanded to serve more students. That's the only possible way that more total people could have gone to college.

Later, US society decided there still weren't enough people going to college. So the US government offered to subsidize insurance on student loans, in order to allow more students to get bank loans to pay for college. And what happened that time… is that US college prices skyrocketed. So now people are leaving college with a crippling, immiserating load of student debt, and if you don't like that, don't go to college.

So: the second time a subsidy was tried, there was some increased barrier or difficulty to starting a new college. Or at least, it was hard to start a new college that the new students actually wanted to go to. We can also infer that the demand function sloped down very slowly with increasing price: lots of people really really wanted to go to college. Or really really wanted to go to a college with an existing reputation, instead of a new for-profit college that nobody had heard about. And existing reputable colleges were unwilling or unable to expand. We know all this, because otherwise the price couldn't have skyrocketed.

You can't effectively subsidize housing unless higher housing prices can cause there to be more housing.

You can't effectively subsidize education unless higher tuitions can cause new attractive universities to spring up, or old reputable universities to expand.

There's literally nothing you can do to cause more people to have more healthcare by moving around insurance premiums, unless the resulting higher prices are causing more doctors and more hospitals to exist.

All this cost disease isn't a simple issue for our civilization, and economists don't agree on exactly what's happening, let alone what to do about it.

On the other hand, politicians almost always only talk about demand -- who purchases the good, how much they pay. They talk about supply not at all. They talk about mortgage tax deductions and federally insured mortgages; restrictions on building apartments, not so much. When costs go up, they talk about the need for higher subsidies instead of asking why supply isn't already expanding. "Where are the new colleges?" they don't cry. "How do we have so few hospitals, with prices so high?" they don't say. "We must do something about the limited number of orthodontic residencies so that more children can have braces!" you will never hear any legislator say. Which is, on standard economics, a recipe for prices that never stop going up.

If you were around for the 2008 Presidential election in the US, there was this a case where some gasoline refineries had broken down, and thus the US was experiencing a gasoline shortage with correspondingly high prices.

And there was floated a proposal to temporarily repeal some gasoline taxes…

…which would have been a pure gift to existing gasoline refineries. The total supply of gasoline was price-unresponsive; it physically couldn't go up until new refineries were built or repaired. The situation was almost perfectly analogous to the city under siege in which there were only 3000 loaves of bread and no more could be made.

Pretty much every academic economist in the United States, Republican and Democrat alike, agreed in unison: "Lifting the gas tax will not change prices at the pump. It won't even cause gas stations to make more money. Literally the only people who benefit from shifting this demand curve to the right are the owners of gasoline refineries."

Presidential candidate John McCain was in favor of temporarily lifting the gas tax. Presidential candidate Hillary Clinton was in favor of temporarily lifting the gas tax. In an extraordinary moment that stunned all of us who knew economics, Barack Obama came out against lifting the tax.

Of course McCain and Clinton almost certainly knew why the measure would be futile. And they probably weren't in the pay of gas refinery owners either. What's going on, I think, is that political journalists believe that voters are too stupid to understand literally any abstraction whatsoever. It doesn't matter whether or not journalists are correct about that, because so long as journalists believe that, they will report on any discussion of economics as a blunder in the political horse race. And politicians know that being reported on as having 'blundered' can be fatal. So McCain and Clinton didn't dare publicly oppose decreased gas taxes, even though they both knew it was stupid. Or something.

I haven't the tiniest idea what to do about that.

The cost disease across housing, education, and above all, medicine, seems like it could all by itself smash an otherwise pleasant outcome where everything gets cheaper as a result of automation.

I have no idea how anyone in AI or machine learning can do anything to help solve this.

Somehow or other, the political equilibrium seems to naturally forbid any politician to mention, at all, what economics would suggest as the actual key elements of the problem.

I have no good ideas on how to solve that either.

What standard economics does say is that you can't solve a cost disease by having the government pay more or trying to move existing subsidies around. You can't help it with a startup that makes doing medical paperwork more efficient. To make costs come down, you need to either (a) make people stop wanting surgeons, bachelors' degrees, and bedrooms; or (b) you need to somehow make more of those things exist, in a world where supply is currently not increasing even as the prices are already skyrocketing.

Absorbed costs

At this point somebody usually points out that various studies are showing that insurance companies, or doctors, or whoever, are not making an excess profit.

There's a phenomenon here I don't fully understand, which looks to me something like this: when a good is in restricted supply and high demand, weeds start to grow on the supply chain.

This isn't just a financial phenomenon. In the US there was a recent controversy about whether medical residents, doctors-in-training, ought to be allowed to work 30-hour-shifts -- 30 hours on the job without a break -- or whether limiting them to 24-hour shifts would be safer. Trying to organize my mental understanding of this phenomenon, it seems to me that this could not happen if not for the fact that there is a huge oversupply of people who want to be doctors, compared to people who are allowed to become doctors. So if you impose this kind of horrible agony, the students don't flee, but stick around.

It's not that anyone is deliberately sitting down to think about how to torture students. It's not even that the medical school has a financial incentive to do it. My guess is that there's a noise, an entropy, a carelessness, that is present in organizations by default; and the only thing that opposes this entropy is if it threatens the organization's survival. So long as the organization can go on operating and making money even if it is torturing its medical residents, there just isn't enough counterforce to oppose the entropy that operates by default to make people's lives horrible.

Similarly with universities and the explosion of administrative costs. If we were thinking of the universities as intelligent beings trying to maximize their profits, they would be opposing administrative expansion regardless of the overall industry situation; every dollar paid to a needless administrator is a dollar out of their own pockets. But that's not actually how large organizations work; they have no such unified will. There's an entropic force that adds administrators and paperwork by default, and so long as the university's survival is not threatened by the present level of entropy, so long as the college goes on getting enough students and tuition and alumni donations to keep functioning, then there isn't the organizational will to oppose that entropy. Most individual people aren't absolutely driven to maximize their earnings and minimize their expenses, so long as they can afford their apartments and not lose their jobs. We should expect this tendency to be even greater for large organizations where no one executive suffers all the organizational inefficiencies as their personal out-of-pocket losses. Why should any particular person drive themselves mad trying to stop it, especially if other parts of the organization are unenthusiastic about supporting the effort? I know little about the empirics of this field, but what little of the literature I've read suggests that in practice, firms seem motivated to cut costs when they must do so to stay competitive -- to survive at all in the market -- more than they automatically do so out of an organizational will to save every possible dollar.

For whatever reason, whether or not the above story is anything like correct, there does seem to be some analogue of Parkinson's Law ("work expands to fill the time available for its completion") which says that in conditions of restricted supply and low competition between suppliers, costs and inconveniences and barnacles expand into the excess wiggle room so created, whether or not anyone profits thereby. Lots of medical residents want medical residencies in restricted supply, so medical residences become horrible to expand into the wiggle room provided by that demand. If university tuitions are skyrocketing because of supply restrictions and subsidies, then administrative costs will expand into that slack. On my hypothesis this is because of background entropic forces that no longer pose threats to organizational survival, but for whatever reason it certainly does seem to happen.

Which implies that the massive amounts of paperwork and administrative costs in the medical industry are not the cause of high prices for medicine, they are caused by high prices for medicine.

Restricted supply and subsidy means the prices must be high to equalize supply and demand. Since they have to be high, and especially since custom makes it look bad to just take out all that money as shareholder profit, there is a vast wiggle room into which will expand barnacles, hospital paperwork, insurance paperwork, high-cost secondary suppliers for goods and services, etcetera. On my hypothesis this is because when you are a supplier in restricted license and demand is high, these inefficiencies do not threaten your organizational survival and so they happen by default. But even if you don't buy that particular hypothesis for why barnacles expand to fill the wiggle room, they clearly do.

So there's no point in trying to fix the price of medicine by trying to eliminate all that inefficient paperwork. It will just grow back as barnacles somewhere else. It never caused the price increase in the first place. There had to be a skyrocketing price to balance the subsidized demand with the restricted supply, and something automatically filled the wiggle room created by that price.

Rising rents

From a speech by Winston Churchill in 1909:

Some years ago in London there was a toll bar on a bridge across the Thames, and all the working people who lived on the south side of the river had to pay a daily toll of one penny for going and returning from their work. The spectacle of these poor people thus mulcted of so large a proportion of their earnings offended the public conscience, and agitation was set on foot, municipal authorities were roused, and at the cost of the taxpayers, the bridge was freed and the toll removed. All those people who used the bridge were saved sixpence a week, but within a very short time rents on the south side of the river were found to have risen about sixpence a week, or the amount of the toll which had been remitted!

Of course this outcome was an inevitable consequence of there being a limited amount of housing on the south side of the river. The landlords couldn't have refused to raise those rents--they could instead have introduced a new hidden price in the time required to apply for apartments, or the social capital and pull required to get the apartments, but supply and demand must equalize. If housing was relatively unrestricted, the higher rent might later induce the construction of more buildings on the same land; but that couldn't happen instantly when the bridge toll dropped.

"Rent" has a number of different complicated definitions in economics, most of which are pretty much equivalent for our purposes. I googled around briefly and picked one that I liked:

"The essence of the conception of rent is the conception of a surplus earned by a particular part of a factor of production over and above the minimum sum necessary to induce it to do its work." (Joan Robinson.)

Land doesn't need any inducement to go on existing and supporting buildings; the supply of land has nothing to do with the price of land. So any part of the price being paid to use a building, which derives just from the price of the land underneath the building, is "rent" in the economic sense. Whatever part of the cost is necessary to induce a janitor to keep the building clean, is not "rent" in the economic sense, even if it shows up in the monthly rent payment in the colloquial sense of rent.

I expect most readers coming in will already have heard of rents and have an idea that rent-seeking is a public choice problem. For purposes of discussing what to do about automation-driven unemployment, especially analyzing notions like the basic income, we are not interested in rent as a generic public choice problem. We are worried about a particular kind of rent that increases to soak up all the benefit whenever we try to help people. Also from Churchill's speech:

In the parish of Southwark, about 350 pounds a year was given away in doles of bread by charitable people in connection with one of the churches. As a consequence of this charity, the competition for small houses and single-room tenements is so great that rents are considerably higher in the parish!

Now imagine that instead of being given bread, they'd been given a basic income. The result wouldn't have been exactly the same--if everyone in the country was getting the basic income, competition for those particular houses would not have been as great. But you can see why this particular kind of rent increase is of particular concern.

The economic definition of a quantity of rent

Consider taxi medallions in New York City, before and after Uber (illegally) busted their (legal) cartel. Now that Uber is around in 2017, there are around 650,000 rides per day (300,000 taxi rides and 350,000 Uber+Lyft) in NYC. In 2010 there were around 450,000 taxi rides per day. The difference between these numbers is due to a previous legal limit on the number of rides; only 13,605 taxi medallions were allowed to exist. These medallions cost on the order of $1 million and were owned by holding companies that extracted huge portions of the taxi fares from the actual taxi drivers.

I expect you are probably already familiar with this overall situation, and I mention it just to exhibit the technical definition of rent.

In the non-medallion-constrained equilibrium, suppose that:

We now restrict the number of taxi medallions to 13,605; an inelastic supply of taxi medallions is now required as a new factor of production for taxi rides. As a result:

Then (\$7-\$4=\$3)/ride will go as rent to the owners of taxi medallions. This rent is derived from control of an inelastic bottleneck on the production of rides. To say that this supply is inelastic is to say that there would be no less of it, on the margins, if the price were marginally less; paying \$2/ride instead of \$3/ride to the medallion owners would not induce some medallions to give up on existing. It is this sense that, by the definition of rent, the rentier is on the margins being paid to do nothing; whatever part of the money is "rent", is by definition not there to induce somebody to do more work and create greater supply.

Imagine that aliens magically agree to fund the New York state government, and New York repeals the state income tax. The people in NYC become richer; they gain more purchasing power; they are willing to pay more for their goods; their demand curve is shifted to the right.

If taxi medallions exist:

If taxi medallions don't exist:

Rent-collectors don't always look like idle rich

Although this essay is supposed to mostly not be about justice and morality, economic rents are an obvious flashpoint for concern about unfair deserts and social parasites. So I note parenthetically that rent passing through a person's hand doesn't mean that person corresponds to the stereotype of an idle rentier lying back and being lazy.

In many cases, the metaphorical taxi cab drivers also own the metaphorical medallions. In these cases part of their pay is the amount that they get as metaphorical taxi drivers for their elastic supply of labor, and part of their pay is the pure rent on controlling an inelastic supply of medallions. But all of it looks to them like they are being paid to do work.

This is what it is like to be an orthodontist in the USA when only 350 orthodontists are allowed to graduate per year. You are still running around all day putting on braces, but that is not where most of the money flowing into your office is really coming from. As an orthodontist you almost certainly won't have the slightest understanding of that; of course you deserve your salary, you work hard all day long!

One also observes that somebody has to be an orthodontist; somebody has to fill out the ranks of those 350 people and make the braces. An orthodontist is only personally guilty of socially destructive behavior if they personally make some choice that supports the limit on graduating 350 orthodontists per year.

Furthermore: The people who actually collect the enormous cash prizes are often dead and buried by the time the present day comes around.

My landlord owns a house in Berkeley. It was an expensive house at the time he bought it; the possible increase in future rents was already baked into the price. He had to take out a loan to buy it. He has to make ongoing payments on those loans. His ownership of the piece of land his house is standing on, is idly generating free rents; but my landlord is not seeing any of that money. He does not get free profits and an easy life.

In turn, the bank that gave my landlord a loan is getting some real interest on the actual investment, and some amount of rent in virtue of it being one of a limited number of entities that are legally allowed to be a bank and make loans to people like my landlord. But somebody who owns shares in the bank did not get those shares for free. And so on.

If Berkeley repealed all the housing restrictions tomorrow, my landlord would be screwed. That's what happened to the owners of taxi medallions in New York City, many of whom took out loans to buy the medallions, when Uber and Lyft came along.

When the original rent-seekers decades earlier talked bureaucrats into creating supply restrictions (for the good of the dear people who must be protected from bad suppliers, of course!) a new necessary factor of production was created. Decades later, people must take out loans to buy that factor of production; and so they make no excess profit. There are no twenty-dollar bills lying in the street; there is no easy way to become an idle rentier and never have to work again.

Observe that these people now have a strong incentive to keep the supply restrictions in place.

This is technically termed "rent-seeking." But that term sounds like we're talking about Elsevier grabbing all the academic journals and charging monopolistic rent for sitting back and doing nothing, where before journal subscriptions were cheap. Of course Elsevier is not unique; there are plenty of villainous rent-seekers trying to actively tighten supply restrictions, or building monopolistic fences around goods that are cheap to produce. But in practice, "rent-seeking" often also comes from people who aren't seeing any excess profit themselves and would be completely screwed over if the supply restriction were busted.

As it happens my landlord is a libertarian, pardon me, neoliberal. So far as I know, he has never personally voted to maintain a housing restriction. But I wouldn't see it as Elsevier-style mustache-twirling villainy if he did.

Monopolistic rents and the Ferguson Police Department

Before the city of Ferguson became a national flashpoint due to Michael Brown being shot by the local police department, Ferguson had 32,975 outstanding arrest warrants for nonviolent offenses, in a town of 21,000 residents. (By comparison Boston, with 645,000 people, issued 2,300 criminal warrants.)

Bluntly, the residents of Ferguson were being treated as cattle and milked for fines.

What happens if you try to give the residents of Ferguson a basic income?

Well, if the citizens of Ferguson were previously being milked for around as much as they can output… why wouldn't the Ferguson Police Department just issue more warrants, once the citizens can stand some more milking?

Stepping back for a moment and considering some less charged technical definitions, "monopolistic rents" arise when a single price-setter controls all of some factor of production; as if a single cartel owned all the taxi medallions in NYC. Then they can collect even higher rents, in some cases, by setting the supply lower than the maximum that factor of production allows. Maybe if you set the price of a ride at \$10, all but 300,000 taxi riders drop out of the market, who can be serviced by fewer taxi drivers, and those who remain are those willing to work for \$3.50/ride. Then the total rent you collect is (\$6.50 * 300,000), higher than the (\$3 * 450,000) you would receive if all medallions were being used.

This is mustache-twirling villainy: the gains from trade on 150,000 taxi rides are being destroyed in a socially negative-sum game.

(There are non-economists who imagine that this alone is the entire business of some evil force labeled "capitalism", but let's not go there.)

Elsevier collects monopolistic rents on its captive journals. There are other science journals in existence, but researchers tell the university that they need particular journals that Elsevier controls, and the university cannot substitute other journals instead.

If you've been following along with the rest of this document, you may have previously wondering something like, "According to this overall outlook on price, if \$2,000/year is the supply-demand equalizing price on a journal, why blame Elsevier for charging that much?" The answer is that the marginal cost of production for a journal is very low; if lots of people were competing to supply a particular journal, more universities would have that journal and the price would be much lower. Since Elsevier has total control of the supply of a good that costs them very little to produce, they can make the supply curve be anything they like; and so they really are solely to blame for the point where that supply curve intersects demand.

For purposes of talking about hypothetical productivity-driven unemployment and remedies like the basic income, monopolistic rents especially matter because they can rise to consume any increase in productivity or purchasing power. If you have total control of a necessary factor of production, if you can at will say "no" and shut down the entire trade, you can charge whatever tariff you like on that trade. You can try to capture as much money as the trade can stand without shutting down; take nearly all of the gains from trade for yourself. If the trade becomes more gainful, you can just grab more of the gains, and all the other traders are no better off.

We could try to shoehorn the Ferguson Police Department into this view, by saying that they have a monopoly on "not being in jail" and that this is a necessary factor of production for which they can demand any price they like. But we don't actually need an economic view of the Ferguson PD; the point I mean to convey is that the Ferguson Police Department can take whatever they want from you, and the more you can afford, they more they can take.

Now consider what happens if there's more than one person who has sole control of one of your factors of production.

In this case, everyone predating on you faces a kind of commons problem. They all want you to survive to go on being milked, so they don't want the milking collective to take too much; but if they demand any less money themselves, that just leaves you with more money for a different milker to seize.

However this situation resolves itself, it's really not going to help if you try to give that person a basic income. It doesn't even help much to pry one of the milkers off their back, if there's at least two more poorly coordinated milkers remaining.

I sometimes go around saying, "Reading essays written by actual poor people in the US suggests that the first thing we could do to help them is to outlaw towing fees." But even this is just a gloss on an even worse situation: maybe it doesn't help to outlaw \$100/day towing fees so long as that just makes court costs and late payment costs go up somewhere else.

Now consider the present system of intellectual property rights. In particular, patents.

The problem with this system is not just that the US patent office goes around issuing patents on "a system that uses the letter 'e' in online communications" or whatever.

The problem is that it creates many parties each of whom has the theoretical ability to individually say "no" to, and shut down the production of, any good or service complicated enough to depend on more than one patent.

If you do not have compulsory patent licensing with court-set fees, then why should any one patent troll--or even the holder of a rare real patent--stop short of demanding the company's entire profit?

Bigger companies and a decline of competition

Economists generally expect there to be a story behind a monopoly rent: a barrier to entry, or a barrier to price competition, that prevents anybody else from strolling in and selling similar or substitutable goods more cheaply.

A number of alarming indicators seem to indicate that developed economies and the United States in particular are exhibiting something like stagnant, locked-up markets; fewer startups, fewer successful startups, more goods being sold into markets where there is less competition. This also goes along with other alarming indicators like people moving between states less often, but for now let's focus on the increasing lack of competition; those are the trends that most obviously threaten to keep prices high despite automation, or extract any increases in income.

If you are a libertarian, pardon me, neoliberal, you will loudly observe that quite often the barrier to entry for decreased competition takes the form of a law, since merchants are quite good at crossing obstacles like mere mountains and rivers. Such barriers indeed commonly arise from local or national laws:

The libertarian will then observe the existence of a "regulatory ratchet" in which the volume of law and bureaucracy seems to only increase over time within a country (or, for that matter, an individual large corporation); and suggest that this will be a force for decreased competition.

Conventional economics says that this is all obviously qualitatively correct, and that only the quantitative degree to which it is the key force responsibly for an increasing lack of competition and dynamic turnover could reasonably be disputed.

Other classical forces producing large corporations are economies of scale and network effects. The ways in which regulatory burdens produce large corporations per se, and not just more expensive goods, are just because of the economies of scale and network effects in dealing with laws and regulators.

All fixed costs or up-front costs imply economies of scale, and advanced technology often has a fixed support cost or a large up-front cost. Larger markets in which it's easier to sell to all the consumers, likewise imply "economies of scale" in this sense: you pay a one-time cost in time and effort to set up with Amazon, and then you get access to Amazon's whole market. Whoever has the cheapest price on a standardized good might capture nearly all of Amazon's market within a nationality, which is the winner-take-all special case of economies of scale. To the extent that more goods are supplied through Amazon and fewer through regional malls, that much concentration of the market will emerge without regulatory forces.

(Legal monopolies a la patent rights and copyrights would be filed by libertarians under "Whether or not you think those laws are good ideas, they are certainly instances of the government being responsible for the largeness of the corporation.")

There also other, weirder factors that might possibly be producing increasingly noncompetitive markets:

Eyeballing this landscape myself, it seems to me that there's a lot of force to the libertarian-pardon-me-neoliberal thesis which suggests that a pretty large amount of this non-competitiveness phenomenon would somehow go away if we could, e.g., diminish regulatory and litigatory burdens by a factor of 10 and reform intellectual property rights. Arguendo, regulatory barriers to entry are what initially create an equilibrium where only one, two, or a small handful of companies sell the goods you need. Then once the market is already controlled by large companies, other factors like coordinated pricing can become a problem for consumers, and big corporate bureaucracies become a new barrier to entry for any related companies that need to interact with the big players.

But so far as principle goes, there are forces listed that have nothing obvious to do with the government, such as index fund shareholders. There are forces creating large corporations that aren't just about regulatory barriers, like economies of scale. I can't think of anything offhand I've read that presents a well-researched case about the quantitative extent to which all these forces are "the problem". Eyeballing the landscape, it seems to me that there are a lot more entrepreneurial dogs not barking which are silenced by a law or regulatory compliance burden or threat of litigation, than are being silenced by Google and Facebook being big enormous companies; but this kind of eyeballing is not a substitute for careful investigation.

If you hear anyone mentioning that not all inequality is bad, they are, hopefully, referring to the point that some inequality comes from manufacturing economies of scale that have nothing to do with regulators, and winner-take-all markets that are being won by lowest prices or best goods. At least in isolation, and not considering knock-on effects, these inequality-producing forces are positive-sum and generally beneficial. Other inequality is produced by monopolistic rent extraction that ultimately derives from regulatory barriers or other declines in price competition, which is a negative-sum phenomenon. The good reason to be alarmed by rising inequality is if it's coming from negative-sum rent extraction driven by decreasing price competition. The statistics on dynamism suggest that this may in fact be the case in an increasingly large slice of the economy.

And again to restate the main point: this in-practice empirical trend toward more and more economic interactions being with larger and larger big corporations that stay around for longer and longer, is not just a problem because of inequality per se or a less dynamic society. It's a problem because it increases the extent to which (a) technological productivity increases are unlikely to decrease prices, and (b) any increased income is liable to be extracted in the form of higher prices elsewhere.

All the technology in San Francisco has not made computer programmers there nearly as much better off as one might naively expect from comparing their nominal salaries to Montana salaries. A huge class of non-computer-programmers ended up actively worse off than before Google came into their lives, mostly because of skyrocketing apartment rents. Today, increased productivity is being converted into gloom by housing restrictions; but tomorrow it could be decreasing competition and an increasing prevalence of effective monopolies and duopolies.

Labor markets failing to clear

Markets are said to clear when all the matched buyers and sellers who would be willing to trade at a mutually agreeable price, are actually trading.

Clearing a market doesn't happen automatically. Clearing a market that wasn't previously clearing can be a huge innovation and sometimes even a profitable one. Craigslist didn't literally clear the market for everyone willing in principle to sell old laptops or buy old laptops, nor did Ebay, but they moved the market closer to clearing and sparked a lot of trades that didn't happen before.

If I am shouting "I'd love to sell an apple for 40 cents!" and you are shouting "I'd love to buy an apple for 40 cents!" and the two of us are separated by a gaping chasm in the Earth that prevents us from ever meeting one another, we can say this apple market has failed to clear.

Of course one could also say that we're merely unwilling to pay the non-monetary prices of climbing down and up the chasm. But at that rate, you might as well say that when it's illegal to sell apples for less than fifty cents, we're merely unwilling to pay the non-monetary price of risking jail. So to make the definition non-tautologous, and allow us to talk about markets sometimes not clearing, we shouldn't consider it mandatory to use all-embracing definitions of price.

What about sales tax? If there's a city sales tax of two cents an apple, you are effectively buying the apple for 42 cents (from you) and I am effectively selling it for 40 cents (to me). We can't sell/buy for 41 cents, even if that's a mutually agreeable price. It might seem like smuggling libertarianism in sideways to declare that things like sales taxes are not allowed to count as a legitimate part of the price of an apple. Why not say that the apple comes packaged with the extra good of complying with local laws and supporting your city government? Someone has to pay for the police that prevent you from being outright murdered, so why engage in the fantasy of an entirely taxless environment etcetera etcetera.

I'm not sure how academically standard the following stance is, but:

It seems to me that the notion of a 'clearing market' and what exactly counts as a 'price', is a flexible point of view; we can change the variable definitions and get a different but still reasonable answer. For example, there are people on Silk Road 2 who are willing to accept a risk of arrest as part of the price of buying and selling drugs. If you define prices as being allowed to include the risk of arrest, we can talk about how close Silk Road 2 comes to clearing that market. If we then tilt our head the other way and see the 'market' as the people who would buy and sell drugs at purely nominal prices if that were legal and carried no risk of arrest, Silk Road 2 isn't remotely close to clearing that market so defined.

For purposes of considering technological unemployment and what we will consider to be 'clearing the labor market', I think it makes a lot of sense to factor out literally everything from the prices that could possibly be factored out. Not just minimum wage laws, not just sales taxes, but even things like income taxes and associated paperwork. If I would trade an hour of babysitting for 10 apples, and you would sell me 10 apples for an hour of babysitting, but we aren't willing to trade if you need a business license and I have to work an extra half-hour because of income taxes, I think it makes sense to view this conceptually as a trade that could in principle clear, but isn't clearing. %%note: If you assume that we're considering markets clearing to always be good things (false: consider the market for nonconsensual sex), it would be an aggressive moral stance even from the standpoint of big-L Libertarianism for me to define a labor market as not clearing because of income tax and claim that this already establishes a problem that must be fixed by any means necessary.

I however am not advocating this as a moral stance in which we assume it must be desirable for the market to completely clear and that every possible means to that end must be taken. I think that defining the clearing of the labor market in the most aggressive possible way, is a conceptually useful way to organize our thoughts about potential consequences of technological unemployment and policy responses. %%

Because: If the labor force participation rate is already dropping like a stone, and some people are making not-outright-stupid predictions of a huge new tidal wave of automation coming in, then you should be willing to consider a lot of options, cast a very wide net for policies to analyze. And this includes looking at possibilities like e.g. decreasing payroll taxes or sales taxes on humans who are having trouble trading their labor for apples. So it makes sense to take a step back and look at everything that contributes to or interferes with "the labor market clearing" in the broadest possible sense: literally everyone who'd be willing to trade babysitting and apples if there were literally no other obstacles in the way.

Labor mobility

One of the things that can prevent a labor market from clearing is if we are both willing to trade with one another, but you are at point A and I am at point B, and it is hard for either of us to move.

Once again, skyrocketing rents

Okay, yes, I know, you're probably a little sick of hearing about it by now, but once again:

If I want to trade my babysitting for your apples, and you live in San Francisco, and I live in Montana, and I can't afford to move to San Francisco, we can view this as a labor market failing to clear.

This is conceptually a different problem from rents being extracted from the people who actually do live in San Francisco. We are looking here at an entirely different source of lost value, the trades that don't occur because people can't afford to live near San Francisco at all.

I've seen estimates on "how much would be gained in the US if people could afford to move to where all the new jobs are" in the range of 5-10% of US GDP, and if anything I suspect that's a severe understatement; not to mention that these gains would flow disproportionately to people who are now disadvantaged.

(No, really, those anti-housing laws are a big damn problem.)

Other problems that are extra-bad problems because they also inhibit labor mobility

International labor mobility

For people whose circle of concern does not stop at national borders:

Estimated economic gains if all your fellow Earth humans could easily move to anywhere on Earth where jobs for humans can be found: 50% of planetary GDP.

Again, pretty high on the list of points one ought to ponder if you otherwise expect mass disemployment all over Earth.

This is one place where there's an obvious way that the AI and machine learning community in particular can try to intervene to make things better: develop better automatic language translation. Build apps for those allegedly approaching augmented reality headsets that provide subtitles for speakers, or automatically translate visible text. This will make it easier for people to relocate across national boundaries in seach of jobs, in the cases where that is legal; and some jobs and some labor can more easily move to whatever few countries make them simultaneously welcome. Governments are not the only forces that ever prevent people from doing things; problems and inconveniences like language barriers also count. %%note: Be warned: nationalists will take offense that you are making it easier to trade with non-national people much poorer than themselves. They will almost certainly see it in terms of you benefiting the rich corporations inside their country, at the expense of relatively poor people inside their country whom the rich corporations would otherwise need to trade with instead. My model of their mental model is not so much that they're evil or uncaring, as that their emotions do not believe on a core level that the extremely poor people being benefited actually exist. %% %%note: Be warned: many leftists do not believe emotionally that anyone can possibly be benefiting from a trade where they are still poor at the end and the corporation is still rich at the end. Clearly the poor people are being exploited since they are still poor at the end, so clearly a bad exploitative transaction is taking place and the poor people would be better off if this did not happen. It will appear to them that your automatic translators are facilitating these evil transactions.%%


Suppose tomorrow the heads of Google and Facebook and Apple and Amazon and Microsoft and Tesla and Uber and YCombinator came to me and said, "If we were to all act with one accord, is there anything we can do right now to make life better in the United States, without needing to massively reform the whole US government? You're not allowed to talk to us about anything else, just that one problem."

Then I would reply:

"You should all get together and build a new city someplace with low rents, in a state with no state-level anti-housing laws; and try to organize an understanding and commitment among the new citizens of this city to not pass any laws against building as many skyscrapers as needed. You should move as much of your companies there as you can manage, and as much of the tech industry as you can persuade to follow you, all at the same time. You should put body cameras on the police officers, and not have Mafia-run towing companies with late fees; you should insure your poorer citizens for 7 days of Uber if their cars break down, and let nothing stand in the way of cheap babysitting. You should build a university there, with serious prestige because of all the prestigious researchers you will bring there; and make sure that university is ready to take in as many students as can possibly managed, at tuitions not far from the real cost of teaching, even if that means not having giant LED boards in the athletic centers."

"And," I would continue, "you should locate this new city in a swing state, the largest swing state you can find such that you think this new population will be able to threaten to tilt the vote there in election years. Then your faction will not have zero political power the way it does within California and New York where your votes are worthless, and you can actually start pushing on changes like NGDP level targeting that would help with all the other problems."

After another moment's thought, I would add: "Your first priority on the state level should be unclogging the supply lines on medicine within that state; get as close as you can to outright occupational delicensing. Push price transparency laws. Build nonprofit hospitals that can use nonprofit H1B visas to bring in doctors from India and the UK: some countries produce excellent doctors that can pass the occupational licensing filter, and have supply pipelines that are less clogged. Remember, in the end, everything you do adds up to nothing for the country as a whole if it all gets sucked up in a nationally increased cost of healthcare because you are using goods in limited national supply. Now can I please have a word with you about--"

Honestly, if you want a story about how the tech industry managed to screw over the rest of the United States, locating all the new jobs in a region with anti-housing laws would be number one on my list.

We can even frame a story about how staying in the high-rent regions was selfish, a case of bad inequality in action, a decision for which one might be held morally culpable. When I asked a friend why Google didn't set up a campus outside the Bay Area so it could employ all the programmers who don't want to live near Mountain View, my friend replied that no project manager would want their project to be located outside the Bay Area because then they would be too remote from the center of political power in Mountain View. (I was surprised that court politics dominated Google to that extent, but I checked with other Googler friends and they agreed.)

And once you start looking at it from that angle, you realize that the people who decide Where The Company Shall Be Located, or where the company stays once it's established, are disproportionately people who can afford the rents. Sure, there are network effects to being in New York, there's some real benefit to the company of being there. But there are also many programmers and non-programmers who would be willing to work for less nominal money if they got to have non-cramped houses and sane commutes. Could the company grow faster that way, if it's not one of the companies that really need to be in New York? And I don't know the answer to that. But it's worth observing that the people who make that decision, regardless of what decision would benefit the company the most, are the people who personally benefit the most and suffer the least from being in New York. They personally live in apartments that aren't tiny, they personally get the city amenities, and they're not personally driving the 2-hour commutes.

Of course if I'm not telling a moral story and indulging in some pleasant righteous indignation, then I'd personally chalk up the observed outcome to maybe 5% personal selfishness, tops. In reality it's probably more like 95% network effects where the startup needs to be in easy car range of the venture capitalists; followed by inertial effects where the company grows but it can't move away from the city because employees have already put down roots, or it's hired people with roots. Above all, it's the Nash equilibrium where you do worse by unilaterally moving out of the high-rent region, unless everyone else you're tangled up with moves at the same time.

Nonetheless, for whatever reason it happened, it really screwed over the rest of the USA.

Another way in which the poor states got screwed over is that they're in the same currency region as the rich states, and their wages are anchored to the national minimum wage. Flowing money is required to pay wages; and the national minimum wage means that wages in poor states require minimum amounts of flowing money to animate. When lots of money flows to the rich states, the Federal Reserve estimates that enough money is flowing in the country as a whole, and doesn't want to create any more because then they think there would be too much inflation. Then residents of poor states can't move to the rich states or print more dollars or decrease local wages so that less flowing money can animate each job. The European Union has similar problems with enough euros flowing in Germany and not enough euros flowing in Spain or Italy or Greece.

Can basic income defeat the mysterious poverty equilibrium?

If you'll give me a moment to depart the path of standard economics, I personally have a question regarding a certain bizarre situation, regarding which economics gives no standard answer so far as I know:

Why are there still poor people in developed countries?

A thousand years ago, it used to be that 98% of the population were farmers. Between then and now, agricultural productivity went up by a factor of 100 in developed countries. If you told somebody back in 1017 CE that this would happen, they might naively imagine that there wouldn't be poor people any more. They might naively imagine that very few people, if any, would be forced to work hard from dawn to dusk. Now blink and see the paradox, the bizarre state of affairs: Why was that prediction naive?

Yes, poor people in today's developed countries have nicer shoes. They die less often. They have 20 changes of worn clothing instead of 1. Some of their houses have television sets, which was a great luxury in the 1950s and something that nobody from 1017 CE could have had at any price. They're still poor. Nobody from 1017 would mistake them for being 2017's rich people after five minutes of conversation. They'd be amazed that poor people in 2017 have so much stuff, sure. They would also recognize the hunted haunted looks, the debts bearing down, the desperate scrabble for work, the exhaustion and despair and the towing fees; and they would perceive that these were not the future's equivalent of a thriving, upright farmer with something to be proud of.

If a 100-fold increase in productivity did not manage to give almost everyone at least as much pride as a thriving farmer, can basic income be the last straw that breaks poverty's back? Before we can even begin to answer that, we'd need a good analysis of what the hell happened over the last thousand years that didn't eliminate poverty.

To my present state of personal knowledge, this looks like one of the giant inexplicable mysteries that has been staring us in the face for so long that people forget to be confused by it. By which I mean: consider how, until the early twentieth century, nobody said "Wait, what the hell is syntax and how the hell do human children learn it and go around generating sentences?" In principle, this incredibly deep scientific question could have been asked much earlier, far back in the nineteeth century, by some scientist in search of an important problem on which to found their career. If someone had earlier noticed what they didn't understand, and seen the incredible mystery staring them in the face, in the form of children walking around doing what everyone expected them to do and took for granted was the way that things had always been.

Why the hell doesn't a 100-fold increase in productivity eliminate lives of desperation, despair, exhaustion, hunted looks, hand-to-mouth living, and an unending fear of your car being towed?

I think the existence of a class of people that can't defend themselves from more than one milker, is probably part of the answer. It doesn't seem like nearly a complete answer. I suspect there's also some kind of weird equilibrium in which societies feel freer to destroy more wealth as people would otherwise become richer. People in 1850 didn't give themselves modern levels of regulatory burdens.

But there is some kind of poverty equilibrium, with restoring forces powerful enough to defeat a 100-fold improvement in productivity. I am skeptical that, after the last thousand years, a basic income will finally be the force that defeats poverty once and for all, especially since we don't know why poverty shrugged off all the previous assaults. I wonder if you could actually give everyone in Niger a basic income and actually have everyone in Niger be better off, rather than the village chief and the national government competing for who can seize it first, and the land rents going up, and Monsanto charging more for seeds.

Of course there are these lovely things called "experiments" that mean we can actually try things instead of just theorizing about them. And doing those with basic income still seems like a good idea. I'm just registering my worry that the restoring forces of the poverty equilibrium may not act instantly, and some of them may be dependent on regional rather than local income levels.

So you'd want to test giving the basic income to all the people in a region at once, not one person in a village (this part seems to be getting tested properly in some cases, yay).

But more importantly you'd want to watch out for the people seeming better-off at first, but then a little poorer, and then a little poorer, by the time the experiment ended 5 years later.


Eric Rogstad

If Wisconsin is trading cheese with Ohio, and then Michigan becomes much better at producing cheese, this can harm the economy of Wisconsin. It should not be possible for Wisconsin to be harmed by trading with Michigan unless something weird is going on.

Was "Wisconsin" supposed to be "Ohio" in the second sentence? Or are you contrasting between Wisconsin trading with Ohio and Wisconsin trading with Michigan?