Unencumbered by the prospect of re-election, outgoing presidents tend to use their final speeches to candidly warn against threats they believe to be metastasizing in society. For example, George Washington spoke of the ills of hyper-partisanship and excessive debt. Dwight Eisenhower denounced the waxing power of the “military industrial complex.” President Barack Obama singled out an economic peril in his otherwise doggedly hopeful final address in Chicago: “The next wave of economic dislocations won’t come from overseas,” he said. “It will come from the relentless pace of automation that makes a lot of good, middle-class jobs obsolete.”
Obama articulated a fear felt by many around the world: That all our jobs will eventually be done by robots. Research backs this fear: One study found that automation will threaten at least 47% of jobs in America and up to 85% in the rest of the world. But a number of economists are beginning to argue that this view of automation excludes a lot of the story.
Putting automation in context
To simply argue that automation is going to gobble up jobs ignores the potential for productivity gains. The Business Harvard Review found that the IT revolution led to 0.6% labor productivity growth and 1% of overall growth in Europe, the US, and Japan between 1995 and 2005. “It all hinges on demand,” says Jim Bessen, professor of economics at Boston University. If the productivity gains are enough to significantly boost demand, then job growth may be the result. This is especially true when new technologies create jobs that simply did not exist before, such as social-media managers. In those cases, any jobs created will make a net contribution to the labor market.
Though automation will cost some jobs, it will also create many others. Though automation will cost some jobs, it will also create many others. A case in point is the rollout of ATMs in the US. Introduced in the 1970s, the number of ATMs increased from 100,000 to 400,000 between 1995 and 2010. Running an ATM is cheaper than paying a teller’s salary, so as ATMs became more numerous relative to tellers, the overall cost of each bank branch came down. As it became cheaper to operate a bank branch, more of them opened, ultimately resulting in the number of bank branches increasing by 40% between 1988 and 2004. This means that more tellers were hired to staff these branches, not less. Instead of ATMs putting bank tellers out of work, US bank-teller employment actually increased over the three decades between 1980 and 2010. The reason for the growth was the added productivity gains brought on by automation.
The textile industry is another example of this phenomenon. Despite the fact that 98% of the functions of making materials have now been automated, the number of weaving jobs has increased since the 19th century. As with the ATM example, automation drove the price of cloth down, which increased demand—and eventually caused more job growth.
There is also evidence to suggest that automation can lead to the substitution of one occupation for another. In other words: It’s true that typesetters may no longer be in demand—but graphic designers are. One industry may disappear, but another more specialized occupation might rise in its place. What’s more, jobs that include a high level of automation, like software development and accounting, are growing faster than other jobs in the economy.
The opportunities of partial automation
A lot of the debate around automation ignores the fact that most of it is partial—that not all of the work is taken over by machines. In fact, only one of the 270 occupations listed in the 1950 census has been eliminated thanks to automation: elevator operators.
Fewer than 5% of jobs in the US could be completely automated using current technology. This is an important distinction. If a job is completely automated, then jobs will indeed ultimately be eliminated. But if the process is only partial, employment for that job may in fact increase because of the efficiency gains and possible effects on demand. It’s also worth noting that fewer than 5% of jobs in the US could be completely automated using current technology.
David Autor, professor of economics at MIT, adds that the remaining non-automated tasks “tend to become more valuable.” This is because automation is likely to take over mundane or repetitive tasks, leaving professionals more time to do the things that really require their skills. For instance, automation will help mortgage-loan officers spend less time scouring paperwork when processing loan applications and free them up to issue more mortgages. Similarly, in the sphere of health care, if the diagnosis of most conditions can be automated, emergency rooms could combine triage and diagnosis, letting doctors focus on special cases, increasing the number of patients being treated overall.
This trend is even true in the era of artificial intelligence (AI). In the legal sphere, a bot’s ability to sift through large volumes of legal documents using software during the “discovery” phase of a trial was thought to reduce the number of the legal clerks and paralegals who traditionally performed this role. Instead, by reducing the cost of discovery, automation increased demand for it. The number of paralegals has increased since the introduction of discovery software in 1990.
The limits to automation
At the moment, automation does not appear to be infinite. It is constrained by what economists call Polanyi’s paradox. Named after Karl Polanyi, who in 1966 observed “We know more than we can tell,” the paradox refers to the difficulty in automating an activity that we only understand tacitly: Painting a picture, writing a persuasive argument, or dancing are all tasks that even people who are highly proficient in them are not fully able to describe. We cannot program what we cannot understand. True, there is evidence that machine learning capable of “understanding” such tasks tacitly might eliminate this hurdle, but for the time being, professions that require flexibility and creativity are quite resistant to obsolescence.
In the short to medium term, the main effect of automation will not necessarily be eliminating jobs, but redefining them. As the skills and tasks required in the economy change, our response should not be alarmism or protectionism, but a strategic investment in education—which was, incidentally, one of the last policies Obama pushed for in office.