On Modeling Labor Markets for Fine-Grained Insights

摘要

The labor market consists of job seekers looking for jobs, and job openings waiting for applications. Classical labor market models assume that salary is the primary factor explaining why job-seekers select certain jobs. In practice, job seeker behavior is much more complex and there are other factors that should be considered. In this paper, we therefore propose the Probabilistic Labor Model (PLM) which considers salary satisfaction, topic preference matching, and accessibility as important criteria for job seekers to decide when they apply for jobs. We also determine the user and job latent variables for each criterion and define a graphical model to link the variables to observed applications. The latent variables learned can be subsequently used in downstream applications including job recommendation, labor market analysis, and others. We evaluate the PLM model against other baseline models using two real-world datasets. Our experiments show that PLM outperforms other baseline models in an application prediction task. We also demonstrate how PLM can be effectively used to analyse gender and age differences in major labor market segments.