;
Welcome to the WorkSim project website ! WorkSim is a novel multi-agent tool to study labor markets.
September 2017 - we gave a talk at OECD , NAEC Innovation Lab Workshop [slides]
February 2017 - Our evaluation of the French Labor "El Khomri" law with WorkSim is cited by the European Commission (Country Report France 2017 p. 33)
June 2017 - We presented WorkSim at Oxford University and at Kingston University
July 2016 - Our evaluation of the French Labor "El Khomri" law with WorkSim is cited by the International Monetary Fund (IMF Country Report No. 16/228 p.20- pdf)
Mediapart - Espace de travail - May 26 2016 : Video Interview (+130 000 views on youtube)
France Info - 11 avril 2016 - 9h15 : Interview on French National Radio(audio)
Le Mot de l'éco - France Info - 26 mars 2016
The Conversation - 28 avril 2016: Comment la Loi El Khomri change profondément le marché du travail… mais n’améliore pas l’emploi global - paper reprinted in Marianne, La Tribune and Sud-Ouest
WorkSim is a project created by Jean-Daniel Kant in 2006 to model and simulate the French labour market. It is led by a multidisciplinary team comprising computer scientists (JD Kant, Olivier Goudet) and an economist (Gérard Ballot). It uses techniques derived from artificial intelligence, multi-agent systems, composed of interacting software agents that aim to mimic the behaviour of economic agents. We draw on the most recent microeconomic analysis while taking into account that human rationality is limited in a very complex market.
Unlike many models, WorkSim takes into account both sides of the labour market: individuals and companies. The agents each have specific attributes, determined once and for all at their creation, as well as internal variables that evolve throughout the simulation. Individuals and businesses are initialized on the basis of numerous INSEE data available: demographics (especially the households that are modelled in WorkSim), and economics (employment, contracts, etc.). Thus WorkSim reproduces the labour market on a scale currently varying between 1/2000 and 1/1400 (from 20,000 to 30,000 agents).
The main elements of labour law (CDD and CDI labour contracts, termination indemnities, etc.) are integrated, as well as 3 modules representing: the State recruiting civil servants, JobAds, the equivalent of a jobcenter-type site that lists job offers from companies and job applications from individuals, and a statistical institute that disseminates a limited number of statistical data to agents.
Individuals decide at any given moment whether they remain in their current state (e. g. unemployed, inactive, employed) or whether they decide to change state (e. g. seek a job or go into inactivity for an unemployed person, seek a job in another company for an employee, etc.). Unlike conventional models, these decisions respect the cognitive limitations of human subjects (bounded rationality), information is available but limited in quantity. Individuals look for the state that offers the best satisfaction, among the choices available to them. At the heart of people's behaviour is an extended version of the theory of job search offered in labour economics.
Companies make decisions to increase their profits. Depending on their demand (which varies over time), they decide to create or eliminate positions, hire, evaluate and manage their staff (e. g. promotions). The decision to create a job requires an innovative demand anticipation mechanism, with finite time horizon to take account of limited rationality. It makes it possible to endogenize the choice between a fixed-term contract (and its duration) and a permanent contract.
WorkSim parameters are calibrated automatically (using a powerful optimization algorithm, CMA-ES) to reproduce more than 60 real data (INSEE, DARES, Eurostat,...). In addition to these targets, most of the stylized facts of the French labour market are reproduced by WorkSim (flow, segmentation of the labour market, long-term unemployment, etc.).
We have already evaluated several policies on the labour market with WorkSim: generation contract, reduction of charges, elimination of fixed-term contracts, labour law, etc. For more details, see publications.
See also this presentation.
Here we show the complete simulation cycle in WorkSim. 1 step in the simulation equals 1 week in real life, in order to take into account short contracts.
This figure represents the decisions of individuals, depicted as a State Diagram. At each time step, the individual compares the usefulness of his current state, with the states he can reach, and chooses the best.
Here you will find a selection of some recent publications about the WorkSim project (click on the title to download)
This paper synthetizes our last version of the WorkSim model and its applications to policy evaluation.
Reference: J.-D. Kant, G. Ballot, O. Goudet (2020) – «WorkSim, an agent-based model to study labor markets»
Journal of Artificial Societies and Social Simulation 23 (4) 4
Here we focus on the endogeneous choice of contract type.
Reference: O. Goudet, J.-D. Kant, G. Ballot (2015) – « How to choose a contract type in the French Labor Market ? An agent-based endogenous model », AE 2015 - 11th Artificial Economics conference, Porto.
Reference: O. Goudet, J. D. Kant, G. Ballot (2016) – “WorkSim - a calibrated agent-based model of the labor market accounting for workers' stocks and gross flows”. Computational Economics, July 2016, pp. 1-48. doi:10.1007/s10614-016-9577-0
Project Leader
Associaté Professor wirh Habilitation (HDR) in computer science at Sorbonne University - Sciences
Multi-Agents Systems Team (SMA) of the Paris 6 Computer Science Laboratory (LIP6)
Professor emeritus of Economics at the Panthéon-Assas University
Centre for Research in Economics and Law (CRED)
Former PhD Student at LIP6 - UPMC
Email (clck to send): Jean-Daniel[dot]Kant[at]lip6.fr
Phone +33 1 44 27 88 05
Sorbonne Université - Sciences
LIP6 (Laboratoire Informatique de Paris 6)
DESIR / SMA
4, place Jussieu
75005 Paris