Relational Employment Instability and Wage Penalties: Employment Trajectories as Text

• In Event: Advances in Computational Methods (Co-Sponsored by Mathematical Sociology Section)

Abstract

Employment instability is central to research on social stratification, yet existing measures treat it as an absolute attribute of individuals---a quantity of disruption evaluated independent of context. This paper argues that employment instability is also relational: its consequences depend on how far a trajectory deviates from the prevailing employment pattern within a cohort. Using large language model embeddings to represent employment histories as text, I introduce trajectory atypicality---distance from the cohort-specific centroid in embedding space---as a measure of relational employment instability. Analyzing NLSY79 and NLSY97 data, I find that aggregate employment trajectory distributions are similar across cohorts, but this stability masks a gender-specific reorganization: women's employment trajectories converge toward the cohort center while men's dispersion increases. Wage penalties for absolute employment instability remain stable and gender-symmetric across cohorts. By contrast, penalties for relational employment instability are sharply gender-differentiated in the older cohort but converge in the younger cohort as men's penalties intensify to match women's. This study shows that relational employment instability thus constitutes a distinct and increasingly consequential axis of wage stratification---one that existing measures do not capture.

Author

• Nanum Jeon, University of California-Los Angeles