Supervision : Jean-Daniel KANT

Impacts of digital automation on labor markets : an agent-based approach

NumJobs is a agent-based model which aims to reproduce the labor market, and analyze the effects of automation on jobs and firms. To do so, we build on an existing model, WorkSim, which we extend to implement the consumption of goods, demography, an existing referential of skills, tasks, and jobs, a digital sector, as well as a simplified finance sector. The population in the model represents the French population in 2016 at a scale of 1/931. This model contains two distinct types of agents : individuals and firms. Individuals aim to maximize their state utility each period, which depends mainly on their income and free-time. They live inside households, which interact with firms on the consumption good market. On the other hand, firms aim to respond as efficiently as possible to an evolving demand of their produced good. The core of the model is the interactions between workers and firms, which rely on the skills possessed by individuals and the tasks that firm need to realize for their production. A very large and diverse set of data was analyzed and used to build this model. The labor market was built on ROME, Pole-Emploi's referential of skills, tasks, and jobs, was was initialized based on labor surveys (Enquete Emploi) and job seekers databases (Pole-Emploi). The result is a population of heterogeneous agents with real skills, jobs, employed in firms of different NACE sectors and producing three types of goods (consumption, intermediate, and capital). Demographic surveys were used to initialize a population as close as possible to the real French population. Automation is implemented with digital capital goods which are produced by firms, and are bought by firms to either augment or replace workers on a task. We study the effects of automation on a micro and macro level, with two different experiments, to analyze its effects on the economy and on jobs.

2019-2022 Publications

• 2022
  • N. De Bufala : “Impacts of digital automation on labor markets : an agent-based approach”, thesis, phd defence 05/16/2022, supervision Kant, Jean-Daniel (2022)
• 2019
  • N. De Bufala, J.‑D. Kant : “An Evolutionary Approach to Find Optimal Policies with an Agent-Based Simulation”, Proceedings of the 18^th International Conference on Autonomous Agents and Multiagent Systems(AAMAS 2019),, Montreal, Canada (2019)