Essays on Automation and Work: Skills, Wages, and Employment

Technological advancements, particularly in automation and artificial intelligence (AI), have ignited widespread debate. These innovations inspire optimism for their potential to boost productivity and drive economic growth. At the same time, they provoke concern because they can automate tasks, potentially displacing workers and exacerbating inequality. Despite strong public interest in understanding the impact of automation, concrete evidence remains limited. On the one hand, automation can substitute for workers, negatively affecting employment and wages. On the other hand, it can complement workers, leading to positive outcomes by increasing productivity and firm competitiveness. Skilled workers—especially those with expertise in emerging technologies—play a pivotal role in adopting and diffusing these innovations, and they may benefit as firms compete for their talent through higher wages. The labor‑market implications of automation have grown even more complex with recent advances in digital technologies such as AI. Unlike earlier waves of technological change—such as robotics, which mainly automated routine, low‑skill tasks—modern digital technologies can perform complex, non‑routine tasks. This capability represents a major departure from traditional views of automation and its impact on the workforce. A nuanced understanding of these evolving dynamics is critical as AI and automation continue to gain traction and redefine the nature of work.

This thesis consists of three independent chapters on the effects of automation on wages, skill demand, and employment. Each chapter contributes to the wider literature on how advanced technologies shape labor‑market outcomes.

Chapter 1 examines the earnings effects associated with AI skills—a growing area of interest because these skills are central to firms’ adoption and implementation of AI. I investigate what the observed effects reveal about demand for, and scarcity of, workers with these skills, as well as the underlying reasons for firm behavior. As with earlier technologies, AI is attractive because of its potential to raise productivity. To adopt AI and fully realize its benefits, firms need workers who can develop and implement the technology. Given their pivotal role, such workers are expected to earn higher wages as firms compete to attract and retain them. Reports consistently note that a shortage of AI‑skilled workers is a key barrier to adoption. To address these questions, I combine skill requirements extracted from job‑vacancy data with register data on the individuals hired. This lets me identify who is hired into AI‑related positions and control for potential confounders. I find that AI‑skilled workers earn more, largely because AI use is concentrated in high‑paying firms. When I distinguish between less complex AI users and advanced AI developers, the developers earn significantly more—unsurprising, since their expertise is directly tied to adopting and tailoring AI. Because developers typically specialize in machine learning, reskilling existing workers is a less viable option for firms. Interestingly, I find no evidence that firms raise wages to lure AI‑skilled workers; current productivity gains may not yet justify the high cost of hiring them. Still, the observed wage premium appears linked to demand for AI skills, suggesting that future productivity improvements could make the premium more pronounced.

Chapter 2 explores how automation affects workers by examining changes in demand for their skills. Automation alters the tasks employees perform, requiring different skill sets and raising concerns about potential impacts. We provide a nuanced view of how automation reshapes work by identifying two main channels through which skill demand shifts. The first—addressed in prior literature—involves firms changing the composition of occupations they hire. The second, less‑studied channel concerns changes in skill demand within existing occupations. Using Danish job‑vacancy data and firm‑level “lumpy” investments in machinery to pinpoint automation events, we find that firms mainly adjust skill demand by transforming occupations rather than reshuffling their occupational mix. Task replacement driven by automation contributes to workforce polarization: high‑skill occupations become more complex and earn higher wages, whereas low‑skill occupations are simplified, with less need for many job‑specific skills.

Chapter 3 investigates how workforce exposure to AI technologies affects firms’ employment and hiring, focusing on Sweden’s small, open, highly service‑oriented economy. Using Swedish job‑vacancy data, we identify workers hired to develop or use AI (AI workers). We combine this with register data and an index of occupational AI exposure to gauge firm‑level exposure and add survey data on firms’ AI use. Contrary to some earlier studies, we find that AI exposure tends to raise employment in Sweden. To understand why, we look at underlying factors. Firm size matters: Swedish firms are generally smaller than U.S. firms, and smaller firms are more likely to rely on external AI services, while larger firms develop AI internally. This may explain why AI‑exposed Swedish firms employ relatively few dedicated AI workers. Despite the positive effect on labor demand, AI does not directly boost labor productivity; instead, the employment gains appear to stem from improvements to products and services rather than from changes to employees’ tasks.

Together, these chapters provide a comprehensive analysis of how emerging technologies affect workers. They offer new evidence on the interplay among wages, skill demand, and employment in the context of automation and AI. This thesis deepens our understanding of these dynamics and offers valuable insights for policymakers aiming to navigate the challenges and opportunities presented by technological change. Specifically, Chapter 1 sheds light on the workers who drive adoption of new technologies, Chapter 2 shows how automation changes skill demand, and Chapter 3 addresses how AI exposure influences employment and hiring.

Hellsten, M. (2025). Essays on automation and work: Skills, wages, and employment (Doctoral dissertation, Aarhus University, Aarhus BSS, Department of Economics and Business Economics). Aarhus University.