By Ellis Soodak
Siri and Alexa are about to get even more popular. More than 4 billion consumer devices from Google Home to smart refrigerators will have some type of digital assistant by the end of this year. Coupled with robotic advances in everything from health care, to warehousing, to manufacturing, it is clear that automation is a global force that cannot be stopped.
Like the code behind Alexa’s brain, the implications of automation are complex. On net, automation creates more jobs than it destroys. However, factory workers who lose their jobs to machines will find little consolation in knowing that their loss was for the sake of the greater good. To help more people benefit for the era of automation unfolding before us, policymakers should work to use the benefits from automation to address its costs. To do that, we need to take a closer look at not only the gains for automation, but also which industries, regions, and occupations will take the greatest hit.
At the national level, automation increases employment in three major ways: 1) It boosts consumption of goods and services by decreasing production costs; 2) it creates new types of jobs by introducing novel technologies; and 3) it increases demand for existing jobs that make use of goods and services that have become cheaper to produce. Economists David Autor and Anna Salomons conclude that these productivity gains create more jobs than are destroyed.
Even history’s most disruptive technologies have followed this trend. In the move from an agrarian to industrial economy, many scholars predicted that the replacement of manual labor with machines would decrease employment. What they failed to predict, however, was the millions of new jobs created from spillover benefits, such as in the factories building these machines or at the companies insuring them. These new jobs ultimately outnumbered the jobs that were lost.
But the gains and losses of automation are not spread evenly. When job losses are concentrated and gains occur elsewhere, specific industries, communities, and groups experience automation as a net negative. We see this clustering in three ways.
First, within the industry being automated, automation often destroys more jobs than it creates. While the net effect of automation is positive, Autor and Salomons’ survey of 19 countries over 35 years finds that productivity gains within an industry are indeed associated with a decline in employment in that industry.
We’ve seen this before. For example, although engines significantly increased overall employment, the introduction of engine-powered tractors on farms contributed to the decrease of agriculture’s share of U.S. employment from over 40% to under 2% during the 1900s.
Today, we may be seeing a similar transition in industries such as retail and manufacturing. While those industries are becoming more productive, productivity increases mean that fewer workers are needed. But other industries, which benefit from an increase in production, such as warehousing, will require more workers. As some industries grow and others shrink, policymakers should focus on making the transition as painless as possible for workers who may lose their jobs by increasing access to and the effectiveness of retraining programs.
Second, the industries that are the most vulnerable to the negative effects of automation are concentrated within certain regions. Mark Muro of Brookings found that just ten Midwestern and Southern states contain more than half of America’s industrial robots. As expected, according to research by economists Daron Acemoglu and Pascual Restrepo, an increase in the number of industrial robots causes both wages and employment in the surrounding area to decline. However, the adoption of other types of automation, such as computerization, tends to be more spread out. For these other types of automation, Acemoglu and Restrepo find either neutral or positive effects on employment in the surrounding areas.
As automation spreads, not all regions will have the same experience. Regions with high levels of one type of automation, such as the Rust Belt, will likely experience job losses that may be difficult for local governments to address. If concentrated automation can threaten local communities, policymakers should consider targeted assistance to these areas that will be hard hit.
Finally, the jobs most susceptible to automation are routine jobs that are made up of few, repetitive tasks, which tend to be lower- or middle-skill jobs. Non-routine jobs, on the other hand, require interpersonal or critical-thinking skills that are not easily automated. In yet another paper, Autor explains that this distinction causes automation to help high-skill workers (and some low-skill workers, such as housekeepers) to the detriment of low- and medium-skill ones. This goes for industrial robots, too. Economists Georg Graetz and Guy Michaels found that industrial robots decrease the hours worked by both low- and medium-skill workers but have no effect on total hours worked — meaning that if these robots do help some workers in the industries where they’re implemented, they’re helping the highest-skilled and best-paid employees.
Since automation tends to affect routine jobs, we should focus on teaching the skills typical to non-routine jobs. To succeed in today’s economy, workers need personal skills, thinking skills, digital skills, and job-specific skills, as Third Way outlined in a recent report. Policymakers and employers alike should increase investment in programs that teach these skills to ensure that our workforce is prepared for inevitable change.
Although automation helps employment overall, there are winners and losers. If policymakers ignore this reality, many Americans will suffer. But if they are able to design policies that embrace automation’s benefits while remedying its harms, then we just might live in a future where everyone has the opportunity to earn a good life.
Ellis Soodak served as a 2017 summer intern for the Economic Program at Third Way.
