AREOPENING OF SAVINGS, labor shortages continue to worsen. In America, the number of unfilled vacancies at 9.3 million has never been higher. Job openings in Canada are 20% higher than pre-pandemic levels. Even in Europe, slower to come out of post-containment doors, a growing number of employers are complaining about the difficulty of finding staff. Discussions of labor shortages have focused on social protection policy and economic disruption. But the phenomenon has a deeper lesson. This tells us something about the myths of automation.
Economists have confidently claimed that a wave of job-killing robots is sweeping the job market. the IMF says the pandemic “is accelerating a shift in jobs away from sectors more vulnerable to automation.” In a recent co-authored article, Nobel Prize winner Joseph Stiglitz says the additional costs of covid-19 “are accelerating the development and adoption of new technologies to automate human labor.” In testimony to Congress last year, Daron Acemoglu of the Massachusetts Institute of Technology suggested that more companies are “substituting machines for workers.” But can pandemic-induced automation really create an army of surplus workers if employers complain about a deficit?
Economists had good reason to believe that job-killing automation would increase. Recessions often lead companies to adopt more robots, in part because labor becomes more expensive as incomes, but not wages, fall. In the event of a pandemic, bosses have an added incentive to automate tasks, as research conducted by the IMF showed. Robots don’t need to distance themselves socially. They don’t get sick either. Thanks in large part to government stimulus packages, companies have also amassed free cash flow, which they may now be able to deploy on robotics or artificial intelligence software.
Those who believe that automation is accelerating can cite many examples. In Ohio, Lee’s Famous Recipe Chicken, a restaurant chain, has installed automated voice systems to take control behind the wheel. Pittsburgh International Airport recently became the first in the United States to use ultraviolet robots for cleaning. British farmers are bragging about using more and more machines to pick strawberries and kill weeds. The number of reports mentioning both “pandemic” and “automation” is increasing at an annual rate of 25%.
The automation debate is fraught with speculation and anecdotes. It’s light on the evidence. The quote from a prominent asshole to justify the claim that automation was “already” underway included a New York Times article and an article on theoretical microeconomics. According to some research, last year automatable jobs disappeared in great numbers; but it is difficult to disentangle the effect of technological change from the bottlenecks. It is true that America GDP is almost at its pre-pandemic level even though employment is 7 million lower. This, some say, shows that the economy can get by with far fewer people. But that could just mean that productivity per worker has increased, perhaps because of things poorly understood like remote working. Many of those on the sidelines will find jobs as fear of the virus fades and they find something to suit them, in turn increasing production above pre-pandemic levels.
It’s not just labor shortages that are undermining the story of a growing wave of job-killing robots. In the United States, the wages of the lowest paid workers, who are believed to be particularly vulnerable to automation, are increasing faster than average, unlike in the aftermath of the financial crisis. Borrow a methodology from the Federal Reserve Bank of St Louis, The Economist divided the US labor market into “routine” and “non-routine” roles. Routine tasks involve easier-to-learn patterns for robots: for example, entering data or checking goods in a supermarket. For four decades, routine jobs have slowly declined as a share of the total, as robots have improved (see chart).
Stick to the routine
So far, however, the covid-induced slowdown is going against the trend. Had the pre-pandemic rate been maintained, we estimate that in May 2021, routine jobs would have accounted for 40.9% of total employment. In fact, they now represent 41.4%, which means America now has in the order of a million “more” routine jobs than expected. Perhaps the uncertainty over the variants is delaying some investments in robotics. Just installing new machines is also more difficult in a world of travel bans and quarantines. U.S. imports of industrial robots fell 3% in 2020.
Australia may be a better place to look for signs of a wave of job destruction. After a few tough shutdowns, the country has been under fairly loose domestic restrictions for over a year, giving a glimpse of what could be in elsewhere. Adapting the results of a government study in 2015, we assigned 335 occupations (from “hotel and motel managers” to “complementary health therapists”) a score of zero to 100, reflecting how well they seem automatable.
Automable jobs were in relative decline before the pandemic, falling to 57% of the workforce in 2019. The trend has continued, with evidence of an acceleration of covid-19: 55% of Australians are now employed in businesses. vulnerable professions. (We found similar trends in New Zealand.) Yet Australia’s unemployment rate is almost as low as it was before the pandemic. Howls from employers about labor shortages are even louder than in America. Automation is not, it seems, scrapping people economically.
The pessimists might finally be right. But even if they are not, the predictions of a world without work will continue. This is because the lingering fear of running machines is not really the result of a calm analysis of the evidence. This could hardly be the case, when centuries of technological improvement have never led to widespread structural unemployment. Countries with more robots tend to have less unemployment, not more.
Rather, concerns about technological unemployment are an expression of something else. They reflect a deep fascination and fear of technology. And they reflect the concern of many economists to get policymakers to pay more attention to the employment prospects of those with the least marketable skills, who are still the most vulnerable to economic changes and shocks. These are perfectly understandable motivations. But the next time you hear a warning about job killer bots, think twice. ■
All of our pandemic and vaccine related stories can be found on our coronavirus hub. You can also listen to The Jab, our podcast on the race between injections and infections, and find trackers showing the global vaccine rollout, excess deaths by country, and the spread of the virus in Europe and America.
This article appeared in the Finance & economics section of the print edition under the title “Stay of execution”