A1093
Title: Machine learning for labor market matching
Authors: Sabrina Muehlbauer - Institute for Employment Research (Germany) [presenting]
Enzo Weber - University of Regensburg and Institute for Employment Research (Germany)
Abstract: A large-scale application is developed to improve the labour market matching process with model- and algorithm-based statistical methods. Matching is defined as a job seeker entering employment by being matched with a specific job. Extensive data is used on employment biographies covering individual and job-related information on employees in Germany. The probability of a job seeker being employed in a certain occupational field is estimated. Thus, a list with any number of job recommendations can be produced for each individual person. The main goal is to improve individual matching by using statistical methods. For this purpose, predictions are made using logit, ordinary least squares regression (OLS), random forest (RF) and k-nearest-neighbours (kNN). The findings suggest that ML performs best regarding the out-of-sample classification error, especially RF. Further, an estimation sample using all spells of persons starting employment performs better than an estimation sample containing only transitions from unemployment into employment. In terms of the unemployment rate, hypothetically, the advantage of ML compared to the common statistical methods could make a difference of 0.3 percentage points.
