Over the next 15 years, 2 to 3 million Americans who drive for a living – truckers, bus drivers and cabbies – will be replaced by self-driving vehicles, according to a December 2016 White House report on the ascent of artificial intelligence (AI). An estimate by the University of Oxford and Citi, a bank, predicts that 77 percent of Chinese jobs are at risk of automation over roughly the same period.
Millions of people around the world would lose their jobs under these scenarios, potentially sparking mass social unrest and upheaval.
Yet mechanization has always been a feature of modern economies. For example, while American steel output remained roughly even between 1962 and 2005, the industry shed about 75 percent of its workforce, or 400,000 employees, according to a 2015 paper in the American Economic Review. Since 1990, the United States has lost 30 percent (5.5 million) of its manufacturing jobs while manufacturing output has grown 148 percent, according to data from the Federal Reserve Bank of St. Louis (see the chart below).
Machines are besting humans in more and more tasks; thanks to technology, fewer Americans make more stuff in less time. But today economists debate not whether machines are changing the workplace and making us more efficient — they certainly are — but whether the result is a net loss of jobs. The figures above may look dire. But compare the number of manufacturing jobs and total jobs in the chart below. Since 1990, the total non-farm workforce has grown 33 percent, more than accounting for the manufacturing jobs lost.
As we look ahead to a world populated by smart machines that can learn ever more complex tasks, economists agree that retraining people will be required. And — as is the case with global free trade — any big economic shift creates both winners and losers. What they don’t agree on is the degree to which machines will replace people.
Occupations and tasks: Quantifying jobs lost
Robots are easier to manage than people, Hardee’s CEO Andrew Puzder (Donald Trump’s original pick for labor secretary) said in 2016: “They’re always polite, they always upsell, they never take a vacation, they never show up late, there’s never a slip-and-fall, or an age, sex, or race discrimination case.”
According to the 2016 White House report, between 9 and 47 percent of American jobs could be made irrelevant by machines in the next two decades; most of those positions — like jobs at Hardee’s — demand little training.
The 47 percent figure comes from a widely cited 2013 paper by Carl Benedikt Frey and Michael Osborne, both of the University of Oxford. Frey and Osborne ranked 702 jobs based on the “probability of computerization.” Telemarketers, title examiners and hand sewers have a 99 percent chance of being replaced by machines, according to their methodology. Doctors and therapists are the least likely to be supplanted. In the middle, with a 50 percent chance of automatization, are loading machine operators in underground mines, court reporters, and construction workers overseeing installation, maintenance and repair.
In a 2016 paper, the Organization for Economic Cooperation and Development (OECD) — a policy think tank run by 35 rich countries — took a different approach that looks at all the tasks that workers do; taking “account of the heterogeneity of workplace tasks within occupations already strongly reduces the predicted share of jobs that are at a high risk of automation.” The paper found only 9 percent of jobs face high risk of automatization in the U.S. Across all 35 OECD member states, they found a range of 6 to 12 percent facing this high risk of automatization.
Are we living in an era so different than past periods of change? Industrialization gutted the skilled artisan class of the 19th century by automating processes like textile and candle making. The conversion generated so many new jobs that rural people crowded into cities to take factory positions. Over the 20th century, the ratio of farm jobs fell from 40 percent to 2 percent, yet farm productivity swelled. The technical revolution in the late 20th century moved workers from factories to new service-industry jobs.
Frey and Osborne argue that this time is different. New advances in artificial intelligence and mobile robotics mean machines are increasingly able to learn and perform non-routine tasks, such as driving a truck. Job losses will outpace the so-called capitalization effect, whereby new technologies that save time actually create jobs and speed up development, they say. Without the capitalization effect, unemployment rates will reach never-before-seen levels. The only jobs that remain will require workers to address challenges that cannot be addressed by algorithms.
Yet many prominent economists argue that this new age will not be so different than previous technological breakthroughs, that the gains will counter the losses.
Take ATMs, for example. Have they killed jobs? No, the number of bank jobs in the U.S. has increased at a healthy clip since ATMs were introduced, Boston University economist James Bessen showed in 2016: “Why didn’t employment fall? Because the ATM allowed banks to operate branch offices at lower cost; this prompted them to open many more branches … offsetting the erstwhile loss in teller jobs.”
In a 2016 working paper for the National Bureau of Economic Research, Daron Acemoglu and Pascual Restrepo — economists at MIT — describe two effects of automation. The technology increases productivity (think about that growth in steel output with fewer workers). This, in turn, creates a greater demand for workers to perform the more complex tasks that computers cannot handle. But that, Acemoglu and Restrepo say, is countered by a displacement effect – the people who are replaced by machines may not have suitable training to take on these more complicated jobs. As the workforce becomes better trained, wages rise. The authors conclude that “inequality increases during transitions, but the self-correcting forces in our model also limit the increase in inequality over the long-run.”
Unlike Frey and Osborne, Acemoglu and Restrepo believe the pace of job creation will keep ahead of the rate of destruction.
At MIT, David Autor agrees. In a 2015 paper for the Journal of Economic Perspectives, Autor argues that “machines both substitute for and complement” workers. Automation “raises output in ways that lead to higher demand” for workers, “raising the value of the tasks that workers uniquely supply.”
Journalists and newsrooms in the U.S. and Europe are the subject of a 2017 case study by Carl-Gustav Linden of the University of Helsinki, in Finland. Though algorithms are able to perform some of the most routine journalistic tasks, such as writing brief statements on earnings reports and weather forecasts, journalists are not disappearing. Rather, Linden finds “resilience in creative and artisanal jobs.”
Stephen Hawking, the eminent Cambridge physicist, warned in a 2016 op-ed for The Guardian that artificial intelligence will leave “only the most caring, creative or supervisory roles remaining” and “accelerate the already widening economic inequality around the world.”
While such dire predictions are common in the mainstream press, economists urge caution.
“Journalists and even expert commentators tend to overstate the extent of machine substitution for human labor and ignore the strong complementarities between automation and labor that increase productivity, raise earnings, and augment demand for labor,” wrote Autor in his 2015 paper.
This must be addressed, the White House stressed in its 2016 review, with investment in education, in retraining for older workers, and by strengthening the unemployment insurance system during periods of change.
Autor is sanguine about the government’s ability to prepare workers for the high-tech jobs of tomorrow. “The ability of the U.S. education and job training system (both public and private) to produce the kinds of workers who will thrive in these middle-skill jobs of the future can be called into question,” he wrote. “In this and other ways, the issue is not that middle-class workers are doomed by automation and technology, but instead that human capital investment must be at the heart of any long-term strategy.”
Economists are basically unanimous: the jobs of the future will require more education and creative skills. (The last time the U.S. faced such a challenge, in the late 19th century, it invested heavily in high schools for all children.)
Even so, computers appear to be usurping knowledge jobs, too. IBM claims it has designed a computer that is better than a human doctor at diagnosing cancer; in Japan, an insurance company is replacing its underwriters with computers.
Globalization, free trade and robots
Some American politicians often point at free trade deals, specifically with China and Mexico, as job-killing and bad for American workers. But a growing body of research points to machines as the real culprits. For example, a 2015 study published by Ball State University found that between 2000 and 2010, 88 percent of lost manufacturing jobs were taken by robots, while trade was responsible for 13.4 percent of lost jobs.
Machines cost about the same to operate no matter where they are located. If it costs the same to keep a factory in China or Ohio, a firm would probably prefer Ohio. Whether the firm is Chinese or American, in theory there is the rule of law in America to protect its investment. So for journalists, a question is not where these automated workshops of the future will be located. It is where the robots toiling in them will be made.
- Acemoglu, D.; Resrepo, P. “The Race Between Man and Machine: Implications of Technology for Growth, Factor Shares and Employment.” NBER Working Paper, 2016.
- Arntz, Melanie; Gregory, Terry; Zierahn, Ulrich. “The Risk of Automation for Jobs in OECD Countries: A Comparative Analysis.” Organization for Economic Cooperation and Development, 2016. doi: 10.1787/5jlz9h56dvq7.
- Autor, David H.; Katz, Lawrence F.; Krueger, Alan B. “Computing Inequality: Have Computers Changed the Labor Market?” The Quarterly Journal of Economics, 1998. doi: 10.1162/003355398555874.
- Autor, David H. “Why Are There Still So Many Jobs? The History and Future of Workplace Automation.” Journal of Economic Perspectives, 2015. doi: 10.1257/jep.29.3.3.
- Bessen, James E. “How Computer Automation Affects Occupations: Technology, Jobs, and Skills.” Boston University School of Law, Law and Economics Working Paper, 2016. doi: 10.2139/ssrn.2690435.
- Collard-Wexler, Allan; De Loecker, Jan. “Reallocation and Technology: Evidence from the U.S. Steel Industry.” American Economic Review, 2015. doi: 10.1257/aer.20130090.
- Executive Office of the President. “Artificial Intelligence, Automation, and the Economy.” White House, 2016.
- Frey, Carl Benedikt; Osborne, Michael. “The Future of Employment: How Susceptible Are Jobs to Computerization?” Oxford Martin School working paper, University of Oxford, 2013.
- Linden, Carl-Gustav. “Decades of Automation in the Newsroom: Why Are There Still So Many Jobs in Journalism?” Digital Journalism, 2017. doi: 10.1080/21670811.2016.1160791.