Like many people, I long worried about the specter of technological unemployment – as machines get smarter and gradually can do the jobs that people do, will we reach a point where machines can do everything people can do?

If and when that happens, we may have a paradise of abundance – machines will make everything we want, without any people needing to work.

But at the same time, how will people get money to buy these things?

I no longer think that’s going to be a problem.

Last week’s Economist has an excellent report on “The return of the machinery question”, which examines the problem from a historical perspective. People, after all, have been thinking about this problem since the start of the industrial revolution.

Despite all the hand-wringing, the economy always seems to generate more jobs than automation displaces.

Now I understand why (others have understood for a long time).

I’ll explain with a simplified economic model.


At any given time, the world (or national) economy has the ability to produce a certain amount of stuff. Food, clothing, houses, cars, entertainment, etc. – all the things we humans want.

How much stuff we can produce at any given time depends on:

  • Labor: How many people exist (each has hands and brain)
  • Capital: How many machines, factories, buildings, etc. we’ve accumulated. How much education people have received, etc.
  • Resources: How much raw materials we can easily get at – metals, chemicals, energy, land, etc.
  • Technology: The ways we know to do things and use the things we have.

Of course how well we use these things matters – we can run factories 24×7 or only 8 hours/day. We can have rules that make us waste time and materials, or incent people to be efficient. We can work long hours, or take lots of vacations.

And over time how much of these things we have changes – population can grow or shrink, resources can be discovered or run out, and we can discover new ways to do things that are better.

But, still, there are limits. At any given time, we can only produce some given amount of stuff.


Also at any given time, there is a (again, roughly) fixed amount of money in the world.

Sure, we can print more, but that doesn’t help. If we can make a billion stuffs each day, and there are a billion dollars of money, then the billion dollars will buy all the stuffs, so 1 stuff for each dollar.

(I said this was simplified.)

But if we print another billion dollars, that doesn’t help. Now there are 2 billion dollars, but only 1 billion stuffs each day. So you need 2 dollars to buy each stuff.  (This is inflation.)


When someone buys something, they give money to some person.

Yes, you can buy a building, which is not a person. But to buy it, you give money to a person (who owned it).

The same is true with everything – whether you’re buying labor or capital or resources or technology – the money goes to people.

There’s no place else it can go, except to people.


So, we can make a given amount of stuff, and the given amount of money will buy that stuff.

And money can only be spent on people.

Now we introduce lots of robots that can do most of the jobs that people do. The robots are cheaper to use than humans, so they get used instead of people.

So now there are people doing nothing, who used to be making stuff.

I just said “the robots are cheaper”. This is key.

Let’s simplify some more and assume we’re still making the same amount of stuff (actually we’ll be making more, which makes things better, but let’s ignore that for now).

So we’re making the same amount of stuff, but the robots are cheaper. That means somebody has extra money left over (probably whoever is using the robots, but that’s not important, as we shall see).

Same amount of stuff. Extra money left over. Which can only be spent on people.

The extra money will eventually get spent.

If it’s spent on stuff made by robots, there is still extra money left over. Because robots don’t get paid, only people do.

Yes, robots cost money, but that money is paid to people – the people who make or own them.

So there is still extra money around. Held by people. They can spend it on more robot stuff as much as they like, but that doesn’t use up the money – it just moves it around to other people.

And since we’re already making as much stuff as before, that means it has to be spent on new stuff made by people, that wasn’t being made before. That’s the only place it can be spent.

So now we’re making more stuff than before. And that new stuff was made by people. And the only people available – who aren’t already busy making the old stuff – are the ones who lost their jobs to robots.

So those are the people who get hired to make the new stuff.

That’s it.

That’s why despite 200 years of worries, technology has never caused mass unemployment.

Because it can’t. There is only so much money around at a given time. If money is saved by cheaper robots, that money gets spent on people.

(Yes, technology has caused, and will cause in the future, temporary problems as people switch jobs. That’s different.)

Finally – Since the robots are cheaper, prices for the stuff they make go down. Which means people can afford to buy more of them.

Which means that people actually get wealthier, in terms of how much stuff they can afford to buy. (They don’t necessarily have more money – they might have more or less – but they can afford more stuff.)

They can’t buy more of the stuff made by people, but they can buy more of the stuff made by robots. Which means stuff made by people is more expensive – more valuable – than stuff made by robots.

Which means the wages of the people who make stuff has gone up, in terms of what the money you pay them will buy.

Up, not down.

Technology makes wages rise. Not decline. Rise.

And that is why we have nice things today, like houses, and medicine, and clean rivers, and airliners, and indoor toilets, and the Internet, that our ancestors with little technology didn’t have.

3 Responses to “How I learned to stop worrying about Technological Unemployment”

  1. Thomas Says:

    Well done. You’ve given a good general explanation of the issue. Here are some further thoughts: People who ultimately will be displaced by robots (i.e., technology) are people with very low skills who must work as nannies, yard men, bus boys, and the like. The wages in such occupations are already very low, and the displacement of low-skilled workers by technology will drive the wages lower. In the absence of government programs aimed at “defeating” poverty, wages would drop until low-skilled workers simply are unable to work at prevailing wage rates given the cost of transportation, housing, etc. Many of them will throw themselves on the mercy of family, friends, and private charitable institutions. More of them (than now) would return to the places south of the border whence they came. So, the relevant labor markets would “clear,” in the jargon of economics. Government programs are the wrench in the works. They subsidize unemployment in various ways (e.g., AFDC, food stamps, mandatory health-insurance coverage to age 26, relaxed standards for claiming disability insurance). These subsidies probably are a major reason that the labor-force participation rate among persons under the age of 55 (males especially) has been dropping since 2000, especially during the Obama presidency. In short, things would work out as you say in the absence of government intervention. Given government intervention, people with jobs who pay taxes will end up subsidizing more and more of the people who are displaced by technology.

  2. Dave Says:


    If the hypothetical that “machines can do everything people can do” comes to pass, then it’s not just low-skilled people who are displaced by the robots – it’s everybody, at every skill level of existing jobs.

    But I think my general analysis remains true – despite that, new jobs will appear.

    At least until the machines are legal people who can receive money on their own behalf, the savings from automation ultimately benefit humans, and those savings can only be spent on other humans.

    The transition will surely happen in stages, but I’m not sure your assumption that low-skill jobs will go first is correct. Most high-skill, high-pay jobs today are for people who process information – managers, decision makers, lawyers, doctors (surgeons excepted), programmers, etc. All of those jobs amount to “knowing which buttons to press on the keyboard, and when”. Those may be more easily automated than jobs involving lots of complex physical manipulations – say, changing pillowcases, or bathing babies.

    But even if your assumption about the order of things is correct, and low-wage jobs do go first, it’s far from clear this means lower living standards for low-wage people. Automation will increase total productivity, which lowers prices. The price drop effect may leave even low-wage people with a higher net standard of living, even if money wages go down.

  3. Thomas Says:

    I suspect that lower prices won’t rescue low-income people, especially those who end up with zero earned income. Only redistribution will rescue them.

    A related post at EconLog, wherein the blogger (a professional economist) declares his agnosticism as to whether technology kills jobs on balance:

Leave a Reply