Annual estimates:

• For employed people, the data come from social declarations made by employers (DADS and DSN), supplemented by data from the Information System on Public Service Employees (Siasp) and those of individual employers.
• For self-employed, the sources used are the files of the Mutualité Sociale Agricole (MSA) for those affiliated to the agricultural regime and the files of the Agence centrale des organismes de sécurité sociale (Acoss) for the others.

Quarterly estimates: the data come from the contribution summary slips (BRC) or nominal social declarations (DSN) made by employers, and payroll files for civil servants.

Monthly

Frequency of data collection: monthly reporting of companies' social declarations and payroll files of the State employees.

Before 2007, Annual Employment Estimates were produced from population censuses, which made it possible to calculate "reference" employment levels (for example: employment at the end of 1989 for the 1990 census, end of 1998 for the 1999 census). Between two censuses, annual employment trends were estimated using various administrative sources, from which growth rates applied to the previous year's employment levels were derived. In doing so, it was assumed that there was constant multi-activity between two censuses.

Starting in 2007, the annual employment estimates process synthesizes every year the various sources on employment, salaried and non-salaried. A certain number of processes are carried out in order to obtain a homogeneous measurement of employment between year N and year N-1 : automatic adjustments and validation of headcounts, of their geographical location and activity classification, manual adjustments and validation as part of an appraisal campaign, by comparing when necessary the data by establishment with the quarterly source "Epure" (on the salaried field). This first vintage made it possible to determine a new "base level" of employment.

Between the 2007 and 2018 vintages, the process was then used to measure changes in employment between two successive years. That is:

• Employment levels at the end of 2007 were considered fixed ("base level");

• Each campaign made it possible to measure the N-1 / N evolution rates of salaried employment by sector of activity A88 and department, and of non-salaried employment by sector A5;

• These rates of change were applied to the results of the previous year ("chaining" from the base level) to calculate the levels of the year N.

Employment levels by employment zone are calculated in a similar way, but are then calibrated to an “enclosing zone,” so as to ensure consistency across the different geographic levels. The breakdown of salaried employment by sex and age group is calculated solely from the Estel data for year N and applied directly to the year-N Employment Estimates, without chain-linking of changes.

The 2019 and 2024 vintages brought two “rebasings” of Annual Estimates: employment levels were realigned with those observed in the statistical sources derived from social declarations. This was because the successive chain-linking of earlier years could have caused a degree of drift, owing to unaccounted-for changes in firms’ reporting behavior (such as when the DSN was introduced) and to methodological improvements in the allocation of jobs by sector of activity or geographic area. In each case, levels prior to the rebasing year were back-cast, that is, recalculated so as to preserve past trends. The published series are therefore always “break-free,” even after rebasing.

In addition, the 2024 production of Annual Estimates extended the geographic scope to include Mayotte, and the coverage now includes employees under the simplified agricultural employment voucher scheme (known as “Tesa”).

All quarterly series released are seasonally adjusted, using the X-13 Arima procedure in JDemetra+ software.

The seasonal adjustment models are reviewed once a year for the first quarter release. The coefficients of these models are re-estimated each quarter.