Research

How does Artificial Intelligence (AI) affect the task content of work, and how do workers adjust to the diffusion of AI in the economy? To answer these important questions, we combine novel patent-based measures of AI and robot exposure with individual survey data on tasks performed on the job and administrative data on worker careers. Like prior studies, we find that robots have reduced routine tasks. In sharp contrast, AI has reduced non-routine abstract tasks like information gathering and increased the demand for `high-level' routine tasks like monitoring processes. These task shifts affect all skill groups, mainly occur within detailed occupations and become stronger over time. While displacement effects are small, workers have responded by switching jobs, often to less exposed industries. We further show that low-skilled workers suffer some wage losses, while wages of high-skilled workers remain unaffected. Finally, we document stronger effects of AI exposure in large firms, occupations more susceptible to automation and for older workers.

We examine preferences for gender diversity among co-workers, using stated-choice experiments with more than 5,000 PhD students and university students in Germany. We uncover a substantial willingness-to-pay (WTP) for gender diversity of up to 5% of earnings on average. Importantly, we find that women have a much higher WTP for gender diversity than men. While the WTP differs by career ambition, competitiveness, and family preferences, we find that gender differences in traits and preferences cannot explain gender differences in the WTP for diversity. Our findings provide an explanation for differential sorting of men and women into jobs based on the share of female co-workers.

We study how the effects of exports on earnings vary across individual workers, depending on a wide range of worker, firm, and job characteristics, and their interactions. To this end, we combine a generalized random forest with an instrumental variable strategy. Analyzing Germany's exports to China and Eastern Europe, we document sharp disparities: workers in the bottom quartile (ranked by the size of the effect) experience little to no earnings gains due to exports, while those in the top quartile see considerable earnings increases. As expected, workers who benefit the most on average are employed in larger firms and have higher skill levels. Importantly, we also find that workers with the largest earnings gains tend to be male, younger, and more specialized in their industry. These factors have received little attention in the previous literature. Finally, we provide evidence that the contribution to overall earnings inequality is smaller than expected.

We use causal machine learning to replicate and extend the study by DiTella and Schargrodsky (2013) on electronic monitoring (EM) and criminal recidivism. Combining a generalized random forest with the original judge IV design, we confirm a negative average effect of EM on rearrest rates. We uncover substantial treatment effect heterogeneity linked to demographics and criminal history, but the generalized random forest indicates that no individuals experience an increased rearrest probability from EM. Our findings highlight the benefits of causal machine learning for policy-relevant questions and suggest that the effectiveness of EM depends on the composition of the target population.