Abstract
I propose the use of state-space methods as a unified econometric framework for studying heterogeneity and dynamics in micropanels (large N, medium T), which are typical of administrative data. I formally study identification and inference in models with pervasive unobservable heterogeneity. I show how to consistently estimate the cross-sectional distributions of unobservables in the system and uncover how such heterogeneity has changed over time. A mild parametric assumption on the standardized error term offers key advantages for identification and estimation, and delivers a flexible and general approach. Armed with this framework, I study the relationship between job polarization and earnings inequality, using a high-quality dataset on UK earnings, the New Earnings Survey Panel Data (NESPD). I analyze how the distributions of unobservables in the earnings process differ across occupations and over time, and separate the role played on inequality by workers' skills, labour market instability, and other types of earnings shocks.
Local Organizers: Vincenzo Scrutinio, Annalisa Loviglio