Recent developments in automation capabilities have raised concerns about future job losses and the reallocation of workers. Previous studies have estimated the probability of automation in different occupations; with these probabilities, they have assessed the impact of automation on employment. However, these studies do not consider second-order effects, that is, the impact unemployed workers in one occupation have on the labour supply in a similar occupation. In this work, we evaluate such second-order effects using a network perspective. We develop a labour market model embedded in a network, where nodes are occupations and edges link occupations that are similar enough for a worker to transition between them. We characterize each node or occupation by its position in the network and its employment and vacancy rates. We then study how workers reallocate after an automation shock that decreases labour demand in some occupation and increases it in others. After discussing the model’s ability to reproduce the Beveridge curve and relating it to standard partial equilibrium models, we study the wedge between labour supply and demand at the occupation level. Our analysis suggests that second-order effects due to reallocation can be crucial to determine unemployment and the pressure on wages at the occupation level. Such outcomes depend not only on an occupation’s risk of being automated, but also on the probability of automation of related occupations and the mobility constraints that the structure of the labour flow network adds.


Research Programmes