Fear that machines will take away work is very old, and so is the hope that they will free people from toil. Looking at earlier waves of automation does not tell us exactly what will happen next, but it shows which patterns tend to repeat and which questions to ask. This level looks at the longer arc, beyond the Luddites and the early factories.

An ancient dream and a reluctant economist

Around 350 BCE Aristotle imagined tools that could work on their own: "if the shuttle would weave and the plectrum touch the lyre without a hand to guide them, chief workmen would not want servants, nor masters slaves." He meant this as an impossible fantasy, which showed how much he took forced labor for granted. Two thousand years later the fantasy began to come true, and economists argued over what it meant.

The British economist David Ricardo first held that machinery helped everyone. Then in 1821, in the third edition of his Principles, he added a new chapter, "On Machinery." Ricardo conceded that machinery could harm the interests of the working class, writing that workers' fear of it "is not founded on prejudice and error, but is conformable to the correct principles of political economy." He still favored machines in the long run, but he admitted that a switch from labor to capital could shrink the demand for workers for a time.

From farm to factory to office

The largest job shift in history happened on the farm. About 40 percent of American workers were in farming in 1900; today it is under 2 percent. Tractors, combines, fertilizer and better seeds let far fewer people grow far more food. Similar shifts happened in every rich country. Yet this did not create permanent mass unemployment. Workers and especially their children moved into factories, then offices, shops, hospitals, schools and other services, often for higher pay. The change was slow enough, spread over two or three generations, that many people moved by choosing a different first job rather than by losing an old one.

Horses were less lucky. The number of horses and mules on American farms peaked around 1920 at over 25 million and then collapsed as engines replaced them. In 1983 the economist Wassily Leontief used this as a warning: Leontief suggested that human workers could one day go the way of the farm horse, losing their role to machines no matter how cheap they became. Critics answer that the comparison breaks down because people, unlike horses, can learn new tasks, invent new wants, vote and own the machines.

Recurring alarms

Waves of worry have come again and again. In 1930 John Maynard Keynes gave wide currency to the idea of technological unemployment, meaning job loss that happens when ways of saving labor are found faster than new uses for labor. He saw it as a temporary stage, and Keynes guessed that his grandchildren might work only about fifteen hours a week, since rising productivity would solve the problem of scarcity. Working hours did fall over the twentieth century, but nowhere near that far. In the early 1960s, fear of "automation" was so strong that President Lyndon Johnson set up a national commission on it. The 1966 report of Johnson's automation commission concluded that technology destroys particular jobs but not the need for work.

Tellers and operators

Two modern cases show that the effect of a machine depends on the details. Automated teller machines spread from the 1970s. Many people predicted the end of the bank teller. Instead, according to the economist James Bessen, the number of American bank tellers rose for decades after ATMs arrived, because ATMs made each branch cheaper to run, so banks opened more branches and shifted tellers toward sales and customer service. Teller numbers did start to fall in the 2010s, as online and mobile banking spread, which is a reminder that the story can change.

Telephone operators had a harder time. From the 1920s, the Bell System replaced human switchboard operators with mechanical switching, city by city. A study by James Feigenbaum and Daniel Gross traced what happened to the women who held these jobs. Young women entering the job market found other work, especially in growing clerical jobs, so overall employment did not fall. But older, experienced operators bore the heaviest cost of switchboard automation: they were more likely to leave paid work entirely or take lower-paid jobs.

Worked example: reading the history fairly

A newspaper columnist writes: "Every time people panicked about machines, they were wrong. Jobs always came back. Stop worrying." Another writes: "Machines destroyed the livelihoods of millions. This time will be worse." Test both against this level. The first columnist is right about the long run in the aggregate: total employment in rich countries kept growing through every wave. But the operators show that "jobs came back" can be true for the economy while false for the particular people displaced, who often never recovered their old standing. The second columnist is right about the pain but must explain why the farm, ATM and switchboard cases did not end in permanent mass joblessness. A careful conclusion asks three questions about any new technology: who loses, who gains, and how fast the change comes compared with how fast people can adjust.