WiP Seminar: Vito Stefano Bramante

Abstract

I study the job search outcomes proposed by a recommender system trained on a quasi-random matching protocol in the form of an urn ball meeting technology. The latter is the most prominent probabilistic model used to microfound the aggregate matching function, commonly assumed exogenous in Diamond-Mortensen-Pissarides types of equilibrium models. These simulations are a preliminary yet necessary first step toward the introduction of a recommender system in a search and matching model. The recommender systems used are of the collaborative filtering type. They rely on matrix factorization techniques to predict the labor market outcome.