This paper presents new evidence challenging traditional views on specialization. Analyzing data from GitHub, covering 35 million task allocations across 64,400 software development teams over seven years, I find a negative link between organizational specialization and key productivity metrics, such as output quality, quantity, and user issue resolution time. To demonstrate causality, I leverage the introduction of an automatic task assignment feature on GitHub that evenly distributes tasks among certain team members. After adopting this feature, teams show reduced specialization and improvements in output quality and quantity,along with an increase in team communication. These results highlight a trade-off in non-routine production: while specialization increase task-specific skills, it hinders vital cross-task knowledge transfer essential for innovation.

We study the consequence of political polarization along educational lines in the United States. Descriptively, we show that college graduates are now well to the left of non-college voters on economic and social issues and much more so than 15 years ago. We then estimate the causal effect of a Republican governor on college graduates’ inter-state migration rates, finding that conservative governance substantially reduces the inflow of college-educated workers. Finally, we analyze a structural model of migration that quantifies the implications of plausible changes in political control for cross-state spillovers and college/non-college earnings inequality.