Here's a startling fact: In the 45 years since the introduction of the automated teller machine, those vending machines that dispense cash, the number of human bank tellers employed in the United States has roughly doubled, from about a quarter of a million to a half a million. A quarter of a million in 1970 to about a half a million today, with 100,000 added since the year 2000.
These facts, revealed in a recent book by Boston University economist James Bessen, raise an intriguing question: What are all those tellers doing, and why hasn't automation eliminated their employment by now? If you think about it, many of the great inventions of the last 200 years were designed to replace human labor. Tractors were developed to substitute mechanical power for human physical toil. Assembly lines were engineered to replace inconsistent human handiwork with machine perfection. Computers were programmed to swap out error-prone, inconsistent human calculation with digital perfection. These inventions have worked. We no longer dig ditches by hand, pound tools out of wrought iron or do bookkeeping using actual books. And yet, the fraction of US adults employed in the labor market is higher now in 2016 than it was 125 years ago, in 1890, and it's risen in just about every decade in the intervening 125 years.
This poses a paradox. Our machines increasingly do our work for us. Why doesn't this make our labor redundant and our skills obsolete? Why are there still so many jobs?
I'm going to try to answer that question tonight, and along the way, I'm going to tell you what this means for the future of work and the challenges that automation does and does not pose for our society.
Why are there so many jobs? There are actually two fundamental economic principles at stake. One has to do with human genius and creativity. The other has to do with human insatiability, or greed, if you like. I'm going to call the first of these the O-ring principle, and it determines the type of work that we do. The second principle is the never-get-enough principle, and it determines how many jobs there actually are.
Let's start with the O-ring. ATMs, automated teller machines, had two countervailing effects on bank teller employment. As you would expect, they replaced a lot of teller tasks. The number of tellers per branch fell by about a third. But banks quickly discovered that it also was cheaper to open new branches, and the number of bank branches increased by about 40 percent in the same time period. The net result was more branches and more tellers. But those tellers were doing somewhat different work. As their routine, cash-handling tasks receded, they became less like checkout clerks and more like salespeople, forging relationships with customers, solving problems and introducing them to new products like credit cards, loans and investments: more tellers doing a more cognitively demanding job. There's a general principle here. Most of the work that we do requires a multiplicity of skills, and brains and brawn, technical expertise and intuitive mastery, perspiration and inspiration in the words of Thomas Edison. In general, automating some subset of those tasks doesn't make the other ones unnecessary. In fact, it makes them more important. It increases their economic value.
Let me give you a stark example. In 1986, the space shuttle Challenger exploded and crashed back down to Earth less than two minutes after takeoff. The cause of that crash, it turned out, was an inexpensive rubber O-ring in the booster rocket that had frozen on the launch pad the night before and failed catastrophically moments after takeoff. In this multibillion dollar enterprise, that simple rubber O-ring made the difference between mission success and the calamitous death of seven astronauts. An ingenious metaphor for this tragic setting is the O-ring production function, named by Harvard economist Michael Kremer after the Challenger disaster. The O-ring production function conceives of the work as a series of interlocking steps, links in a chain. Every one of those links must hold for the mission to succeed. If any of them fails, the mission, or the product or the service, comes crashing down. This precarious situation has a surprisingly positive implication, which is that improvements in the reliability of any one link in the chain increases the value of improving any of the other links. Concretely, if most of the links are brittle and prone to breakage, the fact that your link is not that reliable is not that important. Probably something else will break anyway. But as all the other links become robust and reliable, the importance of your link becomes more essential. In the limit, everything depends upon it. The reason the O-ring was critical to space shuttle Challenger is because everything else worked perfectly. If the Challenger were kind of the space era equivalent of Microsoft Windows 2000—the reliability of the O-ring wouldn't have mattered because the machine would have crashed.
Here's the broader point. In much of the work that we do, we are the O-rings. Yes, ATMs could do certain cash-handling tasks faster and better than tellers, but that didn't make tellers superfluous. It increased the importance of their problem-solving skills and their relationships with customers. The same principle applies if we're building a building, if we're diagnosing and caring for a patient, or if we are teaching a class to a roomful of high schoolers. As our tools improve, technology magnifies our leverage and increases the importance of our expertise and our judgment and our creativity.
And that brings me to the second principle: Never get enough. You may be thinking, OK, O-ring, got it. That says the jobs that people do will be important. They can't be done by machines, but they still need to be done. But that doesn't tell me how many jobs there will need to be. If you think about it, isn't it kind of self-evident that once we get sufficiently productive at something, we've basically worked our way out of a job? In 1900, 40 percent of all US employment was on farms. Today, it's less than two percent. Why are there so few farmers today? It's not because we're eating less.
A century of productivity growth in farming means that now, a couple of million farmers can feed a nation of 320 million. That's amazing progress, but it also means there are only so many O-ring jobs left in farming. So clearly, technology can eliminate jobs. Farming is only one example. There are many others like it. But what's true about a single product or service or industry has never been true about the economy as a whole. Many of the industries in which we now work—health and medicine, finance and insurance, electronics and computing—were tiny or barely existent a century ago. Many of the products that we spend a lot of our money on—air conditioners, sport utility vehicles, computers and mobile devices—were unattainably expensive, or just hadn't been invented a century ago. As automation frees our time, increases the scope of what is possible, we invent new products, new ideas, new services that command our attention, occupy our time and spur consumption. You may think some of these things are frivolous—extreme yoga, adventure tourism, Pokemon GO—and I might agree with you. But people desire these things, and they're willing to work hard for them. The average worker in 2015 wanting to attain the average living standard in 1915 could do so by working just 17 weeks a year, one third of the time. But most people don't choose to do that. They are willing to work hard to harvest the technological bounty that is available to them. Material abundance has never eliminated perceived scarcity. In the words of economist Thorstein Veblen, invention is the mother of necessity.
So if you accept these two principles, the O-ring principle and the never-get-enough principle, then you agree with me. There will be jobs. Does that mean there's nothing to worry about? Automation, employment, robots and jobs—it'll all take care of itself? No. That is not my argument. Automation creates wealth by allowing us to do more work in less time. There is no economic law that says that we will use that wealth well, and that is worth worrying about. Consider two countries, Norway and Saudi Arabia. Both oil-rich nations, it's like they have money spurting out of a hole in the ground.
But they haven't used that wealth equally well to foster human prosperity, human prospering. Norway is a thriving democracy. By and large, its citizens work and play well together. It's typically numbered between first and fourth in rankings of national happiness. Saudi Arabia is an absolute monarchy in which many citizens lack a path for personal advancement. It's typically ranked 35th among nations in happiness, which is low for such a wealthy nation. Just by way of comparison, the US is typically ranked around 12th or 13th. The difference between these two countries is not their wealth and it's not their technology. It's their institutions. Norway has invested to build a society with opportunity and economic mobility. Saudi Arabia has raised living standards while frustrating many other human strivings. Two countries, both wealthy, not equally well off.
And this brings me to the challenge that we face today, the challenge that automation poses for us. The challenge is not that we're running out of work. The US has added 14 million jobs since the depths of the Great Recession. The challenge is that many of those jobs are not good jobs, and many citizens cannot qualify for the good jobs that are being created. Employment growth in the United States and in much of the developed world looks something like a barbell with increasing poundage on either end of the bar. On the one hand, you have high-education, high-wage jobs like doctors and nurses, programmers and engineers, marketing and sales managers. Employment is robust in these jobs, employment growth. Similarly, employment growth is robust in many low-skill, low-education jobs like food service, cleaning, security, home health aids. Simultaneously, employment is shrinking in many middle-education, middle-wage, middle-class jobs, like blue-collar production and operative positions and white-collar clerical and sales positions. The reasons behind this contracting middle are not mysterious. Many of those middle-skill jobs use well-understood rules and procedures that can increasingly be codified in software and executed by computers. The challenge that this phenomenon creates, what economists call employment polarization, is that it knocks out rungs in the economic ladder, shrinks the size of the middle class and threatens to make us a more stratified society. On the one hand, a set of highly paid, highly educated professionals doing interesting work; on the other, a large number of citizens in low-paid jobs whose primary responsibility is to see to the comfort and health of the affluent. That is not my vision of progress, and I doubt that it is yours.
But here is some encouraging news. We have faced equally momentous economic transformations in the past, and we have come through them successfully. In the late 1800s and early 1900s, when automation was eliminating vast numbers of agricultural jobs—remember that tractor?—the farm states faced a threat of mass unemployment, a generation of youth no longer needed on the farm but not prepared for industry. Rising to this challenge, they took the radical step of requiring that their entire youth population remain in school and continue their education to the ripe old age of 16. This was called the high school movement, and it was a radically expensive thing to do. Not only did they have to invest in the schools, but those kids couldn't work at their jobs. It also turned out to be one of the best investments the US made in the 20th century. It gave us the most skilled, the most flexible and the most productive workforce in the world. To see how well this worked, imagine taking the labor force of 1899 and bringing them into the present. Despite their strong backs and good characters, many of them would lack the basic literacy and numeracy skills to do all but the most mundane jobs. Many of them would be unemployable.
What this example highlights is the primacy of our institutions, most especially our schools, in allowing us to reap the harvest of our technological prosperity.
It's foolish to say there's nothing to worry about. Clearly we can get this wrong. If the US had not invested in its schools and in its skills a century ago with the high school movement, we would be a less prosperous, a less mobile and probably a lot less happy society. But it's equally foolish to say that our fates are sealed. That's not decided by the machines. It's not even decided by the market. It's decided by us and by our institutions.
Now, I started this talk with a paradox. Our machines increasingly do our work for us. Why doesn't that make our labor superfluous, our skills redundant? Isn't it obvious that the road to our economic and social hell is paved with our own great inventions?
History has repeatedly offered an answer to that paradox. The first part of the answer is that technology magnifies our leverage, increases the importance, the added value of our expertise, our judgment and our creativity. That's the O-ring. The second part of the answer is our endless inventiveness and bottomless desires means that we never get enough, never get enough. There's always new work to do. Adjusting to the rapid pace of technological change creates real challenges, seen most clearly in our polarized labor market and the threat that it poses to economic mobility. Rising to this challenge is not automatic. It's not costless. It's not easy. But it is feasible. And here is some encouraging news. Because of our amazing productivity, we're rich. Of course we can afford to invest in ourselves and in our children as America did a hundred years ago with the high school movement. Arguably, we can't afford not to.
Now, you may be thinking, Professor Autor has told us a heartwarming tale about the distant past, the recent past, maybe the present, but probably not the future. Because everybody knows that this time is different. Right? Is this time different? Of course this time is different. Every time is different. On numerous occasions in the last 200 years, scholars and activists have raised the alarm that we are running out of work and making ourselves obsolete. For example, the Luddites in the early 1800s; US Secretary of Labor James Davis in the mid-1920s; Nobel Prize-winning economist Wassily Leontief in 1982; and of course, many scholars, pundits, technologists and media figures today.
These predictions strike me as arrogant. These self-proclaimed oracles are in effect saying, "If I can't think of what people will do for work in the future, then you, me and our kids aren't going to think of it either." I don't have the guts to take that bet against human ingenuity. Look, I can't tell you what people are going to do for work a hundred years from now. But the future doesn't hinge on my imagination. If I were a farmer in Iowa in the year 1900, and an economist from the 21st century teleported down to my field and said, "Hey, guess what, farmer Autor! In the next hundred years, agricultural employment is going to fall from 40 percent of all jobs to two percent purely due to rising productivity. What do you think the other 38 percent of workers are going to do?" I would not have said, "Oh, we got this. We'll do app development, radiological medicine, yoga instruction, Bitmoji."
I wouldn't have had a clue. But I hope I would have had the wisdom to say, "Wow, a 95 percent reduction in farm employment with no shortage of food. That's an amazing amount of progress. I hope that humanity finds something remarkable to do with all of that prosperity." And by and large, I would say that it has.
Thank you very much.