What will automation do to the labor market if education quality doesn’t improve? COVID-19 offers a preview

|

This page in:

The effects of coronavirus (COVID-19) on the labor market have been devastating. There have been substantial job losses, and initial labor force surveys find these to be especially high for less educated workers. In developing countries, lockdowns have caused migrations, sometimes significant, from cities back to home provinces and villages where family members are employed in agriculture. Governments have instituted massive wage subsidy programs and income support to mitigate the loss in employment.

Job loss among less educated workers, out-migration of low-skilled urban workers back to the informal agricultural sector, and large government subsidy programs were all outcomes predicted by our research under a worst-case scenario. But our topic wasn’t the labor market effects of a pandemic—it was the effects of automation.

COVID-19 is forcing us to rethink the workplace and global supply chains. Automation constitutes a large part of this thinking.  It has emerged as a solution to everything from treating patients to protecting health care workers, and from re-shoring of production to strengthen domestic supply chains (rightly or wrongly) to enabling social distancing and remote work.  But the pandemic is also providing a glimpse of how automation will affect the labor market unless individuals are able to attain the skills to engage with technology.

A skill-biased pandemic

For those with lower levels of education, COVID-19’s effect on employment has been much more severe than for those with higher levels of education. The U.S. Bureau of Labor Statistics’ April Employment Situation reported not only a historic surge in unemployment, but also a substantial difference in effect between low and higher educated workers. Comparing the unemployment rates between April 2019 and April 2020, the unemployment rate for those with a bachelor’s degree or higher increased from 2.1 to 8.4 percent, while for those with only a high school education it increased from 3.4 to 17.3 percent. In Europe, a recent study estimated that 30 percent of jobs were non-essential but could not be done from home—and that these jobs were held disproportionately by poorly educated workers. Those with less education also tended to fare worse in previous economic crises.

COVID-19: Automation in fast-forward

Under a worst-case scenario, those with automation-substituting skills (i.e., skills that machines can replace) will see their productivity drop as they are replaced by technology. Meanwhile, this substitution effectively increases the productivity of those with automation-complementing skills (the skills that automation relies on). This shift will increase wage inequality and, for developing countries with weak safety nets, lead to an outflow of low-skilled workers back to the agricultural sector. The policy response will need to include massive re-skilling of workers and wealth transfers.

However, our research suggests that these outcomes can be mitigated based on the quality of education available. If education systems are strong enough, they can respond to the demand for workers to acquire automation-complementing skills. If they aren’t strong enough, and the supply of these skills is constrained or less elastic, more wage inequality will result.  In turn, more costly social programs will be needed to offset this inequality, and countries will not be able to take full advantage of the productivity benefits that automation offers.

Which skills are “automation-complementing”? Artificial intelligence experts argue that creativity and social intelligence are areas where humans retain a comparative advantage over machines. Several empirical studies, including our own, have found that school attainment and cognitive skills are important predictors of whether individuals can be employed in occupations that are of lower risk of automation. In other words, the jobs of poorly educated workers are more at risk due to automation just as their employment is now at risk during COVID-19.

This is how the effects of COVID-19 mimic the effects of automation: by creating inequality in the productivity that largely reflect workers’ education level. Of course, the pandemic differs from long-term automation trends in obvious ways. It is (hopefully) short-term, so that re-skilling of the workforce is not an immediate policy consideration; the pandemic is also not increasing the productivity of high-skilled workers in the way that automation would. But COVID-19 may result in a permanent increase in the uptake of automation, as has happened in previous crises; some job losses may be permanent.

A stark warning for education systems

The problem in many countries is that education systems appear unable to provide even basic levels of learning proficiency, let alone the higher order cognitive and complex soft skills needed for workers to complement automation technology. This is especially the case for children and youth from poorer families or households with lower socioeconomic status, from disempowered ethnic minorities or indigenous peoples, and for those streamed at a young age into terminal vocational schooling programs. Adult retraining programs are often at the forefront of discussion about policy responses to automation.  But given that many education systems are unable to provide basic skills, retraining adults may not give them the skills needed to complement automation in the long run. 

COVID-19, however, has demonstrated that many low- and middle-income countries are willing and able to reform their education systems rapidly in response to a crisis. Many are enacting reforms very quickly to adapt to school closures and social distancing, with a focus on education quality and equity. But COVID is also offering a preview of what automation can do to the labor market unless learning outcomes improve drastically in many countries. 

The momentum for rapid reform should continue when the crisis abates; learning outcomes must improve for countries to avoid a future where automation creates a labor market similar to what we’re seeing during the pandemic.  

Authors

Kevin Macdonald

Kevin Macdonald, Consultant, World Bank Group

Join the Conversation

Gustavo Arcia
June 29, 2020

This post is very interesting because it rattles my policy instincts. In principle, I would suspect that automation would disproportionately affect the low paying jobs because--by definition--an automatic process dispenses with manual human input. However, those large numbers of low skilled humans also require mid-level human staff with higher levels of education: foremen, supervisors, mid-level managers and clerical staff, who would also be displaced. By definition, the number of people with low wages would exceed the number of people with higher wages because manufacturing/industrial/commercial processes are pyramidal structures with a few at the top and large numbers at the bottom. Under automation, the pyramidal structure becomes more like an inverted T, with a narrow column of supervisors, and a relatively narrow base of lower paid workers who, chances are, would have a roughly equivalent level of education of foremen and supervisors. My instincts fail when, in thinking about automation, one forgets that the pyramidal nature of labor in a conventional firm, would be transformed into an inverted T, and that this transformation requires a new set of social sector policies. Under increased automation we should pay attention to education and to compensatory programs for those displaced. Currently, we are not paying much attention to this issue. The evidence of a failure to do that is right in front of us: the low income workers in southern states that lost their jobs to China and to local automation got really impacted, and as a result they voted for the candidate offering compensation. A good policy framework in the face of automation should offer compensatory programs aimed at people that are no longer trainable (people 50+ years old, for example) or people facing a dismal job market that is a product of automation. The message then should be that to reduce the human toll of automation we should increase education quality and relevance, and complement it with compensatory programs for the displaced, while the new automated economy creates new types of jobs under a new Inverted T labor paradigm.

July 01, 2020

Thank you Gustavo. These are great insights. It would be interesting to explore the new social programs needed. In my opinion, the time has come to look at Lifelong Learning again.
Thank you,
Harry