Automation, work, and skills: what do we know?

Source Link

Marcus Casey and Sarah Nzau

What do we really know about how technology will impact employment? Which workers will be impacted most, both in terms of class and gender? What role can retraining play? These questions are addressed in a new series of three academic papers on automation published by the Future of the Middle Class Initiative (FMCi) here at Brookings.

Research Assistant - Center for Children and FamiliesConcerns about the impact of technological change on jobs, wages, and the economic security of workers are not new. Most major technological advances cause social disruptions that can often be painful for impacted workers and communities. While the long arc of history has shown that, by and large, past technological change has us wealthier and more productive, it is important to consider both the needs of people potentially harmed in the interim and potential policy choices that can mitigate the harm.

These issues are particularly salient today. Advanced robotics and other automating technologies in concert with the emergence of human-mimicking artificial intelligence (AI) protocols both have the potential to raise the productivity of workers whose skills complement them well, but also to displace workers for whom their skillset competes. While almost all jobs are likely to change to at least some degree, with a reorganization of the tasks contained within them, research suggests that concerns about widespread loss of jobs are overblown. But an important caveat is that many middle – skill jobs that pay decent wages and benefits are particularly at risk of being displaced.

The new papers, discussed at a recent Brookings private seminar, summarize recent developments in the academic literature, suggesting directions for future research and/or surveying options for policy reform or innovation. We summarize these papers below, focusing in particular on their implications for future work.

AUTOMATION AND THE MIDDLE CLASS

Henry Siu and Nir Jaimovich, in “How Automation and Other Forms of IT Affect the Middle Class: Assessing the Estimates,” explore the role of skill-biased technological change. They focus specifically on how advances in the technological capabilities of machinery, equipment, and software are contributing to job polarization, i.e. why employment growth has generally been weighted toward the lower-tail and especially upper-tail of the wage distribution. Siu and Nir Jaimovich discuss existing evidence on the combined contribution of automation, trade, and offshoring on these trends. Their comprehensive review of empirical work shows that job losses to this point have been concentrated in occupations that feature routine tasks, arguing that the occupations least at risk of displacement are those that require human interaction. The Figure 2 from their paper is presented below and highlights the stark differences in job growth across occupations of differing task content.

They also push for more quantitative, policy-oriented research on the consequences of automation and AI for the middle class. While the existing empirical literature quantifies the role of technological change on employment, there is too little information on the potential welfare impacts on workers and families affected by these structural changes in the labor market and potential policy options to mitigate potential harms. Quantitative, policy-oriented models would account for macroeconomic factors such as:

Changes in the occupational employment structure and types of tasks that workers perform

Differing elasticities of substitution among high-, middle-, and low-paying workers in response to automation

Underemployment and changes in labor force participation

Existing redistribution programs aimed at middle class workers

Future policy-oriented research should focus on the role of interpersonal skills, labor market and retraining programs, and the relative role of globalization and automation in driving these employment dynamics. (Read the full paper here).

MEN NOT AT WORK? GENDER AND AUTOMATION

Patricia Cortes and Jessica Pan, in their paper “Gender, Occupational Segregation, and Automation,” note that since tasks vary in terms of their susceptibility to automation, and men and women are typically do different jobs, even within similar occupations, men and women likely face different risks from automation. They study how automation, occupational segregation, and gender gaps in skill acquisition and job transitions interact. A deeper understanding of these trends should enable more directed policy responses aimed at alleviating the distinct challenges that male and female workers may face in a changing economy.

Cortes and Pan propose a new routine-task intensity (RTI) index that measures the susceptibility of an occupation. They use the RTI to investigate how occupational segregation contributed to gender differences in job automation risk between 1980 and 2017. Historically female-dominated occupations typically had a higher risk of automation, but recent changes in the labor market have led to a resorting of women away from those occupations. Figure 2B from their paper, reproduced below, illustrate these trends using their preferred measure. The graph shows that in 1980 occupations in which women were more heavily concentrated were also occupations that ranked high in routine task intensity and, consequently, were at high risk of automation. Since that time, female worker share has dropped substantially in those occupations with high routine task intensity as they have sorted more heavily to occupations in the middle to lower part of the routine task intensity distribution. No similar pattern is observed for men.

Women were more likely than men to be in occupations with a high automation risk in 1980, but between 1980 and 2017, they disproportionately shifted into management-related and medical occupations, with lower risks.

Educational attainment played a key role here. Women were disproportionately more likely to shift into high-skilled occupations because of higher educational attainment; by contrast, men were disproportionately more likely to move into low-skilled occupations. Other factors such as changes in gender norms also contributed to observed changes in employment distribution.

The authors note that trends reported in this paper are not causal, but conclude that today, male-dominated occupations are more exposed to higher automation risk. This is particularly important given that women are increasingly outpacing men in educational attainment and acquisition of skills required to succeed in the labor market of the future—including interpersonal and social skills. These trends suggest that men may be at a distinct disadvantage to women for success in the technology-driven labor market of the future. (Read the full paper here).

TRAINING FOR AN AUTOMATED WORLD

In “Employment and Training for Mature Adults: The Current System and Moving Forward,” Paul Osterman argues that society should better prepare adults for labor market disruption caused by technological change. In particular, technological change contributes to increased risk of job displacement and occupational skill requirements. Citing the fact that 30 percent of adults now work in jobs that pay less than $15 per hour, he notes that technological change is exacerbating inequality. This problem is particularly acute for displaced older workers, as they tend to have lower educational attainment and face potential employers that are reluctant to invest in their training.

Importantly, Osterman argues that the US lacks a coherent training system for adults: that is, a well-articulated and easily accessible set of programs or opportunities that provide pathways for skill acquisition. Rather, it features a “cafeteria approach” that combines community colleges, work force development initiatives, employer training, on-line programs, and apprenticeship programs. Aside from their lack of coherence, uncertainty remains regarding the efficacy of many existing initiatives.

Retraining workers for the modern economy requires a combination of public and private actors to develop best practices, developing appropriate curricula, and expanding access to a broader population. Osterman proposes policies along the lines of:

Increased transparency in job skill requirements

Individual training or lifelong learning accounts

Increased federal funding for workforce development programs

As well as mitigating economic distress due to earnings losses and job displacements, Osterman argues that such policies will also limit spillover effects in communities facing these shocks. The danger, he warns, of ignoring these issues goes beyond individual households, as the prospective changes to work driven by automation and AI may potentially have implications for social stability. (Read the full paper here).

IMPLICATIONS AND FUTURE DIRECTIONS

The key takeaways from this work center on education and skills, and the need to confront changing norms in the labor market. In particular, evidence-based policies aimed at helping workers train into and transition across jobs are key. In addition, this training should be focused on providing better matches between employers and employees. However, these policies to help workers and investments in human capital must take occupational segregation into account since men and women will be impacted by automation differently.