Pernille Plato defends her PhD thesis at the Department of Economics

This Ph.D. dissertation consists of three self-contained chapters. The three chapters study different research questions and apply different methods in doing so, but they share a common focus on how firms and workers navigate through economic transformations.

 

This Ph.D. dissertation was completed as part of the 4+4 Ph.D. program at the Department of Economics, University of Copenhagen. Please note that Chapter 1 builds on and repeats text from the master’s thesis completed as a part of the program.

 

In the first chapter, I study how automation affect firm market power. My findings show that automated firms are larger, more productive, and charge higher markups compared to non-automated firms. I show that this wedge in markups has consistently increased over the past decades. By matching each automating firm to a non-automating firm in the same industry with similar size and growth patterns before the adoption decision, I show that firm markups steadily increase in the years following the first adoption of automation capital. Additionally, I find that automated firms gain market shares from similar non-automated firms within the same industries and demonstrate greater resilience, being less likely to exit the market compared to non-automated competitors. Taken together, these findings suggest that firms with already high markups invest in automation technology to further improve their competitiveness and increase markups. I perform a decomposition of the aggregate markup that underscore the close link between changes in the aggregate markup and automation. The observed initial decline in the aggregate markup primarily stems from non-automated firms, while entry and reallocation f economic activity towards automated firms, that simultaneously are larger, charge higher markups, and charge higher markups as they expand, have pushed the aggregate markup upwards.

 

In the second chapter, we examine how automation decisions by firms affect the health of workers. To do so, we focus on a cohort of manufacturing workers with a strong attachment to the labor market, and their initial employer in particular, before the arrival of industrial robots in the manufacturing sector. We follow these workers’ health records over the next decades, regardless of transitions into unemployment, changing of employers, occupation, industry, etc. At the same time, we are able to track the robot investments of the initial employer through imports of industrial robots in the foreign trade statistics. We find that industrial robots reduce workplace injuries, especially for individuals working in physical occupations prior to the surge of industrial robots in the Danish manufacturing sector. In contrast to the improvements in workplace safety, we find that robots result in more hospitalizations due to stress, depression, and burnout among employees. Opposite to the improvements in workplace safety, many types of employees suffer the distress from robots throughout the organization.

 

In the third and final chapter, we study investments in human capital after loss of physical ability. To do so, we utilize that workplace injuries occur quasi-randomly within occupations. Our first finding is that injured workers enroll in full-time higher education programs after the injury, while short, non-degree training courses as well as basic and vocational degree programs see almost no uptake by injured workers. Exploiting differences in eligibility driven by prior vocational training, we find that higher education moves injured workers from disability benefits to fulltime employment. We find that reskilled workers end up in higher paying, cognitively intensive occupations, and that reskilling shields injured workers from depression. Combining the effects of reskilling on earnings, taxes, and transfers in a cost-benefit framework, we find that reskilling subsidies for injured workers pay for themselves four times over. In the final part of the paper, we evaluate the counterfactual effects of reskilling more injured workers. We find that injured workers reskill based on their idiosyncratic returns, such that expanding the program implies rolling it out to workers with lower returns to reskilling. We use the marginal treatment effects to determine the optimal rates of reskilling for injured workers and find that the current rates are especially sub-optimal for middle-aged workers between.