This paper estimates the causal effect of PhD programs on innovation, exploiting the staggered rollout of doctoral programs across Italian universities. Opening a program increased local patenting by 23% between 1986 and 2001. A shift-share IV based on pre-reform research capacity shows that gains concentrate in technologies close to the fields of local programs. A regression discontinuity design using admission exam scores shows that 27% of the effect reflects increased patenting by program graduates, while most of the remainder reflects spillovers to local firms. Even under conservative assumptions, the value of induced patents exceeds program costs by at least 63%.

Firms underinvest in general relative to specific skills training because workers with transferable skills are more likely to be poached by competitors. This paper studies how labor market competition shapes the type and amount of training firms provide, and quantifies the implications for productivity and welfare. Using novel administrative data from the Italian Fondo Nuove Competenze, we construct a task-based, worker-level measure of the generality of firm-provided training. Comparing workers in the same firm and occupation across local labor markets, we find that a one-standard-deviation increase in competition lowers training generality by about 2 percentage points (5% of the mean). We rationalize this fact with a model in which firms choose the amount and composition of training under poaching risk, and estimate it by GMM on local mobility flows and training choices. The estimated model shows that poaching risk depresses both the amount and the generality of training. Counterfactuals show that restricting worker mobility, in the spirit of non-compete agreements, raises output and welfare because lower poaching risk leads firms to train more in both general and specific skills.

This study examines the impact of digital technology adoption on Italian firms’ performance and workforce composition. Analyzing 2011–2020 data, we focus on six technologies categorized as “information” (e.g., Big Data, Cloud Computing) and “operation” (e.g., Robots, IoT). Employing synthetic difference-in-differences methods, we find that adoption increases sales by 5% and employment by 1% for both technologies. Workforce composition effects vary: information technologies raise the share of older graduates and younger non-graduates, while operation technologies increase young, non-graduate workers without significantly altering the overall educational mix. These findings suggest that digital technology adoption enhances firm productivity but may reinforce existing skill polarization, offering limited benefits to highly educated young workers and potentially reducing opportunities for older, less-educated employees, while operation technology adoption has more muted effects on the skill composition.