Are occupations "bundles of skills"? Identifying latent skill profiles in the labor market using topic modeling

Abstract

Occupations are a central unit for understanding inequality in the labor market, yet we know little about why occupations matter. The existing literature often assumes that occupations are distinct bundles of skills, so skills are rarely conceptualized and measured independently of occupations. How does this limit our understanding of wage inequality? In this article, we use a unique dataset of millions of online job postings in the United Kingdom to capture the skill content of work at the job level and analyze its relationship to existing occupational classifications. While previous literature has often defined different skills as uni-dimensional and independent from each other, we propose a skill profile approach to capture the combination of skills that workers need on the job. Using topic modeling on highly detailed job skill requirements, we identify the skill profiles of job postings and analyze the extent to which they affect wages within or across occupational categories. Our results reveal substantial heterogeneity in skill content within occupations, and show that both job-level skills and occupations are key to explaining wage differentials across jobs. These findings challenge the often assumed role of occupations as bundles of skills, and offer new perspectives for analyzing labor market stratification.