Indicator for the Regional Labor Market Using Machine Learning Techniques: Application to Colombian Cities

Abstract

This article proposes a methodology to estimate a labor market indicator that combines economic, social, inequality, and expectation variables. Machine Learning techniques are used to select the most relevant variables. The indicator captures the traditional evolution of the employment and unemployment rates and incorporates information on gender, age, informality, productive sectors, and Google Trends data. This approach allows for a more comprehensive understanding of the labor market situation, better visibility of regional differences, and analysis of the heterogeneous impact of the pandemic and subsequent recovery. The methodology is exemplified in the Colombian cities of Cali, Medellín, Bogotá D.C., and Popayán.