EMPLOYMENT IN FRANCE IN THE ERA OF BIG DATA

Scientific project

In a society that is organized around the dynamics of its labour market, a better ability to anticipate this dynamic and a closer consideration of the initiating factors becomes more vital than ever.

Although the dynamics of labour market is a key issue, it is difficult to comprehend. Its proper understanding involves considering a variety of factors such as:

• growth of the various sectors of activity
• the attractiveness of different geographical areas
• the skills needed and available
• levels of supply and demand

These variables may be endogenous or exogenous to the labour market, updated daily, monthly or annually, public and available or private and in varying degrees capable of being estimated. However, the historical inability to interconnect, compare and integrate them within short time limits underlines the limitation of current statistical methodologies in providing detailed analysis (geographical, sectoral, etc.) and reliable prospective analysis.

The emergence of technologies related to the Big Data environment  – heterogeneous data comparison, real-time or near real-time access to data from multiple actors, tools for systematic analysis of the correlation between unstructured data, etc. – suggests that new, more powerful, more accurate and  more reliable methodologies are called for. The cross-fertilization of data from multiple sources could also allow the identification of hitherto unknown explanatory factors for labour market developments, or of rules for defining reliable forward-looking analysis in this market. Randstad wants to improve its knowledge of the labour market in order to anticipate its evolution. To this end, it has aggregated data pertaining to demand (CVs), and supply (job advertisements) in the socio-economic context (INSEE data).

The processing of this data base raises two main research questions:

• On the one hand, this database contains a large volume of data whose size increases exponentially due to the regular updating of the information. Moreover, the data is variously structured and unstructured, thus raises challenges with regard to its interpretation. Finally, some data may be missing, thus raising questions of interpolation. Under these circumstances, current statistical techniques are no longer relevant for providing robust responses.
• On the other hand, this data base may prove inadequate for providing a precise assessment of the dynamics of the labour market. Combining current information with macroeconomic sectoral data specific to the employment areas concerned would probably enhance knowledge of the market.

Thus in order to comprehend the dynamics of employment, two research axes will be emphasized:

Axis 1: Data analysis and predictive models

The objectives of this axis are as follows:

• Evaluation of the consistency of structured and unstructured data,
• Estimation of gaps in the data,
• Detection of weak trend signals.

Axis 2: Open Data
The objectives of this axis are as follows:

• Testing to what extent additional data would improve the measurement of employment dynamics,
• Prototyping a collaborative innovation project on the sharing of “employment” data,
• Project experimentation and impact assessment.