Logotyp: till Uppsala universitets webbplats

Essay I:  I study how digital skill requirements affect the gender composition of new hires in female-dominated administrative occupations in Sweden from 2006 to 2016. I track the prevalence of digital skill requirements in Swedish vacancy ads and analyze the gender composition of new hires through a matching process using Swedish register data. I also examine changes in age composition and its interaction with gender. Although men are overrepresented among computer-related educational fields and occupations, I find that the share of women has not decreased, despite a significant increase in digital skill requirements. Using employer fixed effects, I observe an increase in female hires ages 36-50 for ads that specify digital skill requirements. This may be due to the fact that increased exposure to technology increases the complexity of tasks, resulting in a higher demand for skilled workers, which are predominantly women in female-dominated occupations. 

Essay II: I study the effects of robot and software technology exposure on the evolution of the gender wage gap in cities across the US from 1980 to 2010. Technology exposure, measured via task-replacing patents, has reduced labor demand in primarily male-dominated occupations. Consequently, the gender wage gap declined more in cities with high employment shares of exposed occupations. Moving from the 25th to the 75th percentile of city-level software exposure explains approximately 15% of the average decline in the gender wage gap in the sampled cities. The city-level association between robot patent exposure and changes in the gender wage gap is lower and less robust, possibly due to lower average wages for robot-exposed occupations. I perform a tentative analysis of AI-exposed occupations and show that AI patents are mostly related to tasks performed in male-dominated occupations high up in the wage distribution.

Essay III: (with Ola Andersson, Niklas Bengtsson, and Per Engström) Previous research suggests that student evaluations of teacher performance are biased against women. We test this hypothesis on economics students in a randomized, double-blind experiment, set up in a natural educational setting. During the Covid-19 pandemic, teaching assistants moved online and answered questions through email instead of on campus. We randomly assigned a male or female name to the instructions given by the online teachers. Importantly, the teachers actually responding to the questions did not know if they interacted with the students as male or female, which is a novel contribution to the literature. The course evaluation asked students to rate the mentors' helpfulness, knowledge, and response time. The results show no bias against the female mentor in any dimension. Our confidence interval around the zero effect does not overlap the effect sizes reported in highly influential previous studies.