Abstract

This project examines how labor market institutions - particularly codetermination - shape if and how firms implement artificial intelligence and how the associated gains and losses are distributed within the workforce. It uses the DiWaBe II dataset and the Betriebspanel in combination with information from online job postings to precisely identify where and when firms implement AI. Using a matched multi-period difference-in-differences design, the project estimates the effect of AI adoption on employment, wages, productivity, and training - differentiated by the presence of a works council.