Measuring Labor Market Match Quality with Large Language Models New!

We propose a novel method to assess labor market match quality using large language models (LLMs). Our approach tasks LLMs with the role of a human resources specialist to evaluate the fit between an applicant and a job. Our study consists of two parts. First, we validate the LLM approach using an administrative dataset from an online job platform and show that our new measure of match quality correlates positively with traditional measures, yet it can reveal additional insights. We also demonstrate the applicability of the method with survey data, where traditional measures are infeasible because of the limited sample size. Second, we present two applications of the LLM measure of match quality. The first application simulates an audit study to examine how gender disclosure influences LLM assessments, finding that gender information leads LLMs to favor women for traditional female roles. The second application shows how LLMs may contribute to understanding the versatility of majors. We show that majors that can fit a broad range of occupations are unfairly penalized by traditional measures of match quality, which the LLM measure can mitigate.