Céreq workshop: The impact of big data on the job market




The French and Italian experiences. Prospects for Céreq.

The online recruitment market has been undergoing far-reaching changes for several years. Job advertisement sites have been proliferating to the point where they have become an unavoidable port of call for job seekers and recruiters alike. Said to facilitate the matching of supply to demand, digital technologies, it is argued, make the labour market more transparent and adaptable. Without necessarily addressing all the questions that these digital technologies raise with regard to the functioning of the labour market and its forms of regulation, the workshop organised by Céreq on 29 September 2017 provided an opportunity to  report on the changes taking place in France and Italy.


For a good 20 years now, digital technologies have been exerting a profound structuring effect on the functioning of the labour market, due particularly to the widespread use of the Internet to post job advertisements and applications. Online job advertisements now constitute an enormous mass of data, the use of which cannot but generate interest in the world of research. The purpose of the workshop organised by Céreq on 29 September 2017 was precisely to examine the various ways of approaching this complex topic.


The speakers at the workshop

Yannick Fondeur, socio-economist, researcher at Cnam (Lise/CNRS), member of the Centre d’Etudes de l’Emploi et du Travail.

Mario Mezzanzanica, researcher in statistics and quantitative methods, University of Milan – Bicocca (Italy)

Patrick Rousset, statistician, research officer at Céreq (Département Travail Emploi Professionnalisation)

Madalina Olteanu, senior lecturer in applied mathematics at the University of Paris 1 Panthéon-Sorbonne.


The contributions of a socio-historical analysis of the digital labour market

The first approach to be outlined was that adopted by Yannick Fondeur; it is socio-historical in character. Fondeur is interested in the major organisational upheavals that the labour market has been experiencing for two decades as a result of the ever increasing use of the Internet. Drawing on detailed information on the histories of new entrants into that market, and making an analogy with the natural world, he described a complex ‘ecology’ made up of a multitude of new ‘species’ that have inserted themselves between applicants and recruiters; they include job boards (e.g.: Monster, Cadremploi), classified ad sites (e.g.: Leboncoin, Vivastreet), job ad aggregators (e.g.: Indeed, Jobijoba), recruitment management software developers (e.g.: Taleo, Talentsoft), multiple job board posting applications (e.g.: Broadbean, Multiposting) and so on. Over time, each of these systems has devised its own way of positioning itself as an intermediary facilitating the match between job seekers and companies’ needs. And logically enough, the competition between them is intense and transparency is far from being always guaranteed!

Yannick Fondeur emphasised that the landscape of the digital labour market was in a constant state of flux and could well be undergoing major changes in the near future. These changes would be linked in particular to Google’s direct entry into the market, which would be based on recruitment management software it has developed. Another important change would result from the new European regulations on open data, which would generalise the already considerable flows of job advertisements among operators.

All things considered, this journey deep into the digital labour market revealed a complex world that is still characterised by many ‘imperfections’, notably those linked to the dispersion of job advertisements and the syntactical diversity of their contents. Above all, however, this market is far from hegemonic. After all, it accounts for only a proportion of all job advertisements. Surveys of recruitment practices (such as ‘Ofer’) show that informal ‘side roads’ or ‘back channels’, such as unsolicited applications, are still effective  routes into employment alongside  the ‘major highways’ represented by online job advertisements.


The challenge of constructing a job advertisement system adapted to the actors’ requirements: the case of an Italian database

A second way of approaching the digital labour market is to use the available data in order to develop an information system that can be accessed by the various stakeholders in the labour market. This approach was outlined by a pair of Italian researchers from the University of Milan – Bicocca. By agreement with a series of job boards, their multidisciplinary research team has collected more than two million job advertisements since 2013 from the whole of Italy and in this way has built up a credible database without duplicates.

This approach means they are today in a position to offer a database of job advertisements that are updated almost in real time. The original feature of this database is that it was constructed in order to provide answers to questions that four types of actors in and around the labour market commonly ask in the form of easily visualisable data.

The first type of actor is individuals, who may search the database for information on the most important competences required for a given occupation, which may be located in a given area. This competence-based approach is also the one used by employment agencies, which are concerned to identify the applicant profiles most in demand from companies and the competences on which these profiles are based. Local authorities are also able to submit requests in order to identify the jobs most in demand in their areas and to use the information provided to develop the training programmes required to produce the competences necessary in those jobs. Finally, this database is also of interest to companies, which can augment their recruitment procedures by visualising the occupations most in demand in their area of activity and the competences corresponding to them.

Above and beyond these specific uses by the actors, this database can also be used from a more general point of view in order to enhance our knowledge of the ways jobs are changing. For example, it has been used to improve understanding of how and to what extent ‘digital’ skills are gradually permeating the content of certain jobs.

However, the discussion brought to light two limitations linked to the specificity of the job ads that constitute the basic material for the database and of which users should be aware. The first was articulated by Yannick Fondeur, who pointed out that, in drawing up job advertisements, employers are not only describing a given job and the corresponding competences. They are also addressing the labour market, which may lead them to ask for competences that not only correspond to their actual needs but possibly also to additional expectations. It might be imagined, for example, that, in certain occupations, increasing expectations in terms of ‘soft skills’ reflect recruiters’ strategic aims in a context in which ‘hard skills’ are evolving much less quickly.  A second problem to be aware of is linked to the fact that, in drawing on data focused on competences, users are not tackling the question of training head-on, which may lead to difficulties in finding a common language with the world of training providers.  


Using big data in order better to visualise the competences associated with jobs: the birth of an experimental project

This final presentation outlined the initial ideas for a collaborative research project that Céreq is seeking to instigate with a research team from the University of Paris 1, namely the  SAMM (Statistique, Analyse et Modélisation Multidisciplinaire/Multidisciplinary Statistics, Analysis and Modelling) team, which is made up of mathematicians and computer scientists.

Like the approach described above, the aim of this collaborative project is to develop a tool for analysing competences, taking as a starting point databases of job advertisements. The tool will have a dual updating and processing capability. The first of these capabilities will be developed by using the techniques that have become standard in this area (web scrapping, text-mining, clustering, etc.) and developing partnerships with APEC, a French organisation specialising in the employment of managerial staff that maintains a major employment website, and Pôle Emploi, the French government employment service.

As far as the second function is concerned, the aim will be to draw on methodologies that can be used to examine the importance, evolution or emergence of competences and the qualitative changes they undergo. The proposal under consideration is to present the specific intrinsic concepts (distance, dissimilarity, core, etc.) and original modes of visualising occupations and competences in the form of ‘organised charts’. These charts should shed some interesting light on the debate on transferable competences by suggesting distinctions between ‘stable’ and more ‘inconstant’ competences. They might therefore help to shed further light on a competence’s trajectory over time.

However, it is difficult to say anything more at this stage, since this approach is currently only in the planning phase. But this ambitious project is part of a long-term objective to improve our understanding of the mechanisms underlying the automated analysis of large employment, skills and competences databases maintained by job boards or job ad aggregators. (cf. Yannick Fondeur’s presentation above) that have become widespread in recent years, particularly within the Observatories of. From a research perspective, the aim here is to understand how the various modes of processing these data may or may not provide additional or more reliable information on issues related to intermediation in the labour market.

Haut de page