Coronavirus Shows That Fears About Automation Are Overblown The coronavirus pandemic might spur lots of companies to think harder about automation. by James Pethokoukis

The coronavirus pandemic might spur lots of companies to think harder about automation. For instance: Not only might more commerce be online, but more of the future workers in those warehouses and fulfillment centers might be robots. That’s tomorrow, however, not today. If anything, this outbreak has undercut the idea that somehow we are on the verge of a job apocalypse. As journalist Matt Simon writes in Wired: 

This economic catastrophe is blowing up the myth of the worker robot and AI takeover. We’ve been led to believe that a new wave of automation is here, made possible by smarter AI and more sophisticated robots. … The problem will get so bad, argue folks like former presidential candidate Andrew Yang, we’ll need a universal basic income to support our displaced human workers. … Yet our economy still craters without human workers, because the machines are far, far away from matching our intelligence and dexterity.

Yet little has changed for the folks who touted UBI as a key response to the impending robot takeover. They see all that federal cash being deposited in our accounts as a sort of proto-UBI that will get more Americans comfortable with the notion of Washington regularly sending them a check. Still, we should spare a moment to remember their original thesis: The machines are coming for us.

Then again, maybe not. Some of the most interesting work being done on the true employment threat posed by modern automation is being done by Boston University’s James Bessen. And one of the big takeaways from his research is that automation tends to eliminate some of the tasks that constitute a particular job, but not all them. Rarely are entire jobs automated away, although they might change a lot. And even the jobs that disappear might experience an extremely long decline. This from a late 2019 analysis co-written by Bessen:

Consider what happened to the 271 detailed occupations used in the 1950 Census by 2010 (Bessen 2016). Many occupations were eliminated for a variety of reasons, but few of these were automated away. In many cases, demand for the occupational services declined (e.g., boardinghouse keepers); in some cases, demand declined because of technological obsolescence (e.g., telegraph operators). This, however, is not the same as automation. In only one case—elevator operators—can the decline and disappearance of an occupation be largely attributed to automation. Nevertheless, this 60-year period witnessed extensive automation, but it was almost entirely partial automation. 

 Indeed, when examining companies today in the Netherlands, Bessen finds that automating firms tend to grow employment faster than other firms, both before and after the automation event — although “it is entirely possible that [employment] does not rise as fast as it would have otherwise.” And there was pain for workers at the companies that had an automating event. Again from the study:

We study the impact of automation on incumbent workers, those workers who were employed at their firm for three or more years before the automation event. Over the five years following automation, these workers lose, on average, about 11% of one year’s earnings or, in absolute terms, 3,800 euros. These losses could arise from lower wages or from spells of non-employment. In fact, the daily wage rate does not change for these workers. These workers do not appear to experience reduced wage rates either if they stay at their firm or if they move. The lost earnings come from spells of non-employment attributed to automation. These total 18 days per worker on average (for both leavers and stayers) over five years.

 Relatedly, we see an increase in the share of incumbent workers who leave their firms attributable to automation, although we do not know whether these workers were laid off or left of their own choice. During the year of the automation event, about 2% more of the incumbent workers leave by comparison to the control group. Over five years, the cumulative separations is less than 13%. The separations following automation appear to occur as a trickle rather than as a mass layoff. These separating workers are offset by new hires at the firm, although it appears that the new hires do not fully offset the separations during the first years after the automation. However, our focus here is on the impact on incumbent workers regardless whether they are replaced or not. And the evidence shows a significant, although not overwhelming, loss of income and days worked.

In other words, the employment effects are disruptive rather than apocalyptic. The policy challenge is about helping workers transition to new jobs or industries, not permanent unemployment: “[Policy] measures might include re-training, relocation assistance, and temporary income support. On the other hand, existing policies that discourage or hamper worker transitions, such as employee noncompete agreements, could be viewed as problematic.”