The Network Origin of Slow Labor Reallocation

Abstract

How fast do labor markets adjust to technology shocks? This paper introduces a novel network-based framework to model skill frictions between occupations. Using expert data on skills, I construct a network of occupations and find it is sparse, divided in clusters of similar occupations with 'bridge occupations' linking distinct clusters. Leveraging French administrative data, I show that workers transitioning through these 'bridges' move to occupations with higher wages and lower unemployment. Next, I build a tractable model of job search with networked labor markets, and demonstrate that bridge occupations significantly affect reallocation speed, with slow reallocation creating large adjustment costs. I then augment the model with quantitative extensions, leveraging hat-algebra methods to solve counterfactuals without having to estimate large numbers of parameters. Calibrated to French data, the model predicts that robot adoption induces slow reallocation, around 40 quarters, and that this sluggish reallocation reduces welfare gains by approximately 40%- an order of magnitude higher than previous estimates. However, policies targeting bridge occupations can speed-up reallocation, and much more so than policies targeting tight occupations directly. These findings highlight the crucial role of the occupation network in shaping reallocation dynamics and provide new insights for the design of labor market policies.