IIt was meant be a bloodbath. When Covid-19 struck in early 2020, economists warned that a wave of job-killing robots would sweep through the labor market, leading to high and structural unemployment. A prominent economist claimed in a testimony before Congress this fall that employers are “replacing workers with machines.” A paper published by the IMF Early 2021 said such concerns “appear valid.” Surveys of companies suggested they had big plans to invest in artificial intelligence and machine learning.
Your browser supports this
Save time by listening to our audio articles while multitasking
Wonks had every reason to be concerned. Recessions cause many companies’ earnings to fall, but not wages, making workers less affordable. Some previous downturns have led to bursts of job-killing automation that have robbed people of their jobs and left them on the economic scrap heap, at least temporarily. Covid appeared to pose an additional threat to workers. people get sick; robot not. According to research, past pandemics have accelerated automation.
More than two years later, however, it’s difficult to find much evidence of job-killing automation. Instead of workers complaining about a lack of jobs, bosses complain about a shortage of manpower. On the OECD There are unusually high levels of vacancies in the richest country club, even as the recession approaches. In many countries, wages are rising the fastest for the low paid, people thought to be most at risk of losing their job to a robot.
To test the Doomster’s predictions more directly, we looked at job data for America, Australia, and the UK. Based on a methodology developed by the Federal Reserve Bank of St. Louis, we categorized jobs into “routine” and “non-routine” areas. Routine jobs involve repetitive movements that are easier to learn by a machine or computer, theoretically making them more susceptible to automation.
Over time, and particularly during past recessions, the share of routine jobs in the labor force has declined (see chart). But during the pandemic, the rate of decline actually slowed. In the two years leading up to the pandemic, the proportion of jobs that could be automated in Australia fell by 1.8 percentage points. In the two following years they fell by 0.6 percentage points. We find similar trends in the UK, although a recent coding change makes analysis more difficult. America today has a little more routine jobs than one would expect based on pre-pandemic trends.
Economists are now working on theories that should be less prone to failure. Perhaps the remaining routine roles are particularly difficult to automate. Perhaps, in some cases, technology is actually improving rather than hurting workers’ prospects. For now, one simple rule will suffice: the next time you hear a ludicrous prediction about robots and jobs, think twice.
Sign up for more expert analysis of the biggest stories in business, finance and markets talks about moneyour weekly newsletter for subscribers only.
