Automation in the Short and Long-Term

February 21, 2019

Conversations about the impact of automation on workers are usually divided into two types, the short term and the long term. In the short term, people are worried about specific jobs, mainly of low-skilled laborers, being automated. Examples such as transportation, one of the largest sectors for employment in the United States, being automated by self-driving cars, dominate conversations about the short term impact, framing policy discussions. The long term conversation focuses more on the possibility that most jobs will be automated, again with a focus on lower-skilled workers, as there is a belief an underclass will develop that will need more creative social safety provisions from the government. These conversations, however, should not be separate, and insights on how to resolve the worries of the short term can help shape a more prosperous long term.

In a recent article in Nature, researchers from the Leverhulme Centre for the Future of Intelligence argued that there is a need to bridge concerns about the short term and long term consequences of artificial intelligence, of which automation is a major one. They note that the policy decisions taken to aid workers displaced by technologies soon, such as “educational measures to facilitate transferable skills or financial safety nets for those made redundant,” would shape the impact of automation more generally. By changing the makeup of the labor force, the dynamics of the labor market, and the incentives for firms developing and implementing artificial intelligence, predictions of the long term require an understanding of what will happen in the short term. It is often the case, however, that discourse on the short term impacts of automation is misguided, which clouds longer term judgements.

The most pervasive of these judgements is that it is only the low skilled who are going to be affected by automation, while the high skilled will not. However, a recent study led by James Bessen of Boston University School of Law has called this belief into question. The paper presents the first empirical estimate of the worker-level impacts of investments in automation technologies, looking at firms in the Netherlands from 2000-2016. The authors argue that while low skilled workers are more intensely affected, it is high skilled workers who are affected more frequently. While the Netherlands is a European nation, its economy resembles North American nations in key respects, such as the proportion of its workforce in the service sector. While more work needs to be done in applying this pioneering methodology to the data sets of other nations, the findings nonetheless present insights into the dynamics of automation and short term policy responses.

The paper’s methodological innovation is in looking at the impact of automation across all non-financial sectors of the Dutch economy, rather than focusing on technologies. This is important, as practically, firms invest not only in particular technologies, but entire new ecosystems and management processes that help them improve productivity. Research into the economics of artificial intelligence has increased its focus on the “complements,” that improve adoption, from complementary technologies such as 5G and connected devices, to policy innovations, to operational techniques. Researching how firms actually adopt new technologies, rather than the predicted dynamics of a technology in isolation, provides a much more accurate picture of the effects felt by workers.

Looking at automation through this lens, the findings suggest that tenured employees, those with at least three years at a firm, and particularly older workers, are most likely to be displaced as a result of automation investments. This results from the gradual nature of automation, with new hires being selected within the framework of firm investments, and their contracts making them more difficult to fire. The effects on longer-term earnings are also noticeably different, as new hires, should they be laid off, have an easier time finding new jobs, than older workers who have built up firm-specific skills.

Delving deeper, another finding is that more skilled workers, from both new hires and tenured employees, face a greater risk of automation, whereas lower-skilled workers are less likely to be automated. The authors note, however, that higher-skilled workers usually find new jobs easier than lower-skilled workers, so do not face the same long term income pressures. Taken together, this presents a picture of automation in which short term interventions need to focus less on the availability of new jobs, but on better matching. The fear that mass redundancy results from automation is not borne out by the data. As the authors note, “Compared to findings from literature on mass layoffs, however, the effects of automation are more gradual and automation displaces far fewer workers, both at the individual firms and in the workforce overall.”

The concerns for short term policy responses to automation need to focus then on smoothing the transition for workers, rather than a pessimistic view that they are unemployable or that opportunities are not available. This could involve encouraging firms to make investments in retraining their own workers should they foresee automation investments being made, removing barriers to labor mobility such as through untying benefits from specific employers, and removing the disincentives to hire labor that occur through a distortionary taxation system. By focusing on workers in the short term, it is clear that we can improve the opportunities automation can bring them in the long term.

Ryan Khurana is a Catalyst Policy Fellow, Executive Director of the Institute for Advancing Prosperity, and a tech policy fellow at Young Voices.
Catalyst articles by Ryan Khurana