The future of the labor force: higher cognition and more skills

The ongoing automation and digitization of the workforce are causing significant shifts in employment. While some occupations, like insurance underwriters and data entry keyers, are at risk of obsolescence, others, such as blockchain engineers and digital forensics analysts, are emerging. The transformation isn't simply about job displacement; existing jobs are being reshaped, with automation taking over routine tasks and creating new tasks requiring uniquely human capabilities. This research focuses on understanding how automation differentially affects various skill sets and how workers and educational systems can adapt. The traditional human capital approach to education and training, which views skills acquisition as an investment leading to higher wages, is criticized for its assumption of static skills. This study moves beyond a simple wage-based assessment of skills, focusing on job content and the specific tasks required of workers, offering a more nuanced understanding of skill sets and their susceptibility to automation.

Three main perspectives on technology's impact on labor are discussed: deskilling (simplification of tasks leading to lower worker skill), upskilling (technology enhancing worker skill), and reskilling (workers adapting to automation by learning new skills). The study notes the inherent tension between deskilling and upskilling, particularly concerning the routineness of tasks. Previous research has identified two key skill sets: sensory-physical (involving direct physical interaction) and social-cognitive (requiring interaction, problem-solving, and critical thinking). The paper builds on this categorization to investigate how automation affects these skill sets differently. It argues that sensory-physical skills are more easily automated because they involve objective, repeatable actions, while social-cognitive skills, due to their inherent complexity and subjectivity (as captured by Polanyi's paradox), are harder to automate. Algorithm aversion, the lack of trust in machines for subjective tasks, further complicates the automation of social-cognitive skills. The authors propose three hypotheses: (H1) Higher sensory-physical skill requirements lead to greater susceptibility to automation; (H2) Higher social-cognitive skill requirements lead to lower susceptibility to automation; and (H3) A greater number of skills possessed by a worker reduces susceptibility to automation.

This study utilizes panel data from 2003 to 2022 from the O*NET and OEWS databases, linked using the Standard Occupational Classification (SOC) system. O*NET provides data on the importance of 118 workplace skills for 923 occupations, while OEWS provides employment and wage estimates for approximately 830 occupations. The authors carefully mapped these datasets to a unified SOC2010 system, accounting for the different versions of SOC across years. A Box-Cox transformation was applied to the growth rate of employment share to stabilize variance and reduce the influence of outliers. The main regression model investigates the impact of sensory-physical skills scores (Phy), social-cognitive skills scores (Cog), the number of skills mastered by workers (NS), the offshorability index (OI), and wage changes on the growth rate of employment share. Occupational and year fixed effects were controlled for, using heteroskedasticity-robust standard errors. Revealed comparative advantage (RCA) was calculated for each skill in an occupation to account for ubiquitous and over-represented skills. The number of important skills was also calculated and a binary variable (NS) created based on above-average skill number. The offshorability index was constructed based on the importance of face-to-face interaction and on-site job requirements. Robustness checks were conducted to account for potential impacts of the financial crisis and COVID-19 pandemic by limiting the analysis to specific time periods and adjusting for missing data in wage percentiles. The analysis also includes an examination of the skill scores of 150 emerging occupations between 2000 and 2023.

The regression results support the hypotheses. The coefficient for Phy is negative and statistically significant, indicating that higher sensory-physical skill requirements are associated with a decline in employment share. Conversely, the coefficient for Cog is positive and statistically significant, demonstrating that higher social-cognitive skill requirements are associated with an increase in employment share. The coefficient for NS is also positive and statistically significant, confirming that possessing a greater number of skills reduces susceptibility to automation. The offshorability index (OI) shows a positive and statistically significant effect. The analysis of emerging occupations reveals that most new jobs are concentrated in areas requiring higher social-cognitive skills, particularly in computer, engineering, and science occupations and healthcare sectors. Robustness tests using alternative time periods and addressing missing data in wage percentiles largely confirm the main findings. These results consistently show that jobs with lower sensory-physical skills, higher social-cognitive skills, and a diverse skillset are less vulnerable to displacement by automation.

The findings underscore the importance of cognitive skills and versatility in the face of automation. The differential impact of automation on various skill sets highlights the need for strategic adjustments in education and training. While automation may displace workers in some sectors, it also creates opportunities in fields requiring high-level cognitive abilities and adaptability. The results support the argument that focusing on cultivating diverse cognitive skills and fostering flexibility is crucial for individuals navigating the changing labor market. The findings also have implications for national educational systems, emphasizing the need for curricula that prioritize cognitive skills, interdisciplinary learning, and lifelong learning opportunities. The study's findings are particularly relevant given the rapid pace of technological advancements.

This study provides the first empirical evidence demonstrating the differential impact of automation on sensory-physical versus social-cognitive skills. The results highlight the importance of cognitive skills, diverse skillsets, and adaptability in a rapidly changing labor market. Future research could explore the long-term effects of automation on specific industries, delve deeper into the role of algorithm aversion, and investigate the effectiveness of different reskilling and upskilling initiatives.

The study relies on O*NET data, which are based on a methodology involving subjectivity in assessing skill importance. Future research could explore alternative, more objective measures of skill scores. The analysis focuses on the US labor market; further studies are needed to ascertain the generalizability of the findings to other countries and contexts. The effects of the financial crisis and COVID-19 pandemic might have had an effect on the study results.