Networks Workshop - Leonard Bocquet (Cambridge)

Title: The Network Origin of Slow Labour Reallocation

Abstract:

There is growing concern that displaced workers reallocate very slowly following trade or technological shocks, but still little is known regarding the underlying determinants of reallocation speed. In this paper, I adopt a network-based approach to model the frictions to worker mobility between labor markets, and I show that this perspective offers new insights into the factors influencing the pace of reallocation. First, I leverage French administrative data on skills to construct a skill network, where occupations are nodes, and the connections between them represent potential transitions. The skill network is sparse, but divided in tightly-connected clusters, and features network bottlenecks, which are occupations acting as bridges between those clusters. Second, I build a tractable model of job search with heterogeneous occupations connected by a skill network. The key new ingredient is that workers not only seek jobs within their own occupation but also search in adjacent occupations within the network. Drawing from recent advances on the theory distribution dynamics, I introduce a novel metric to quantify worker reallocation speed. My principal analytical finding stresses the granular impact of bottleneck occupations on the speed of reallocation. Third, I estimate the model's parameters using French matched employer-employee data. I prove that the main parameters can be identified and estimated efficiently, despite their large number, by running reduced-form regressions with structural interpretation. This work has important implications for public policy, e.g. regarding the targeting of employment subsidies and training programs.