French court admits onomastic proof of discrimination

Posted by

The Cour de Cassation is the highest court of law in France. A judgment of the Court of Cassation of December 14, 2022 recognizes patronymic analysis – statistics on the ethno-cultural origin of names – as ex post proof of discrimination in hiring.

This decision was noted and commented on by Etienne Allais, founder of the Entre-Autre firm specializing in supporting companies, communities and organizations in their actions to fight against discrimination and in the management of diversity:

This decision is not a real surprise, because patronymic analysis has already been used for several years for the purpose of measuring inequalities of origin and the method has been “formalized” by the Defender of Rights and the CNIL. It is therefore a confirmation by the Court of Cassation of the possibility of using these analyzes to prevent or demonstrate discriminatory practices. On the other hand, it shows employers that it is in their interest to carry out these analyzes themselves in order to prevent inequalities of origin, at the risk of finding themselves in the same situation as the company in this case.

Etienne Alais, Entre-Autre

Can the analysis of the origin of names (surnames) be used as evidence to prove discrimination in hiring?

It is possible to use the analysis of the origin of names as evidence in an employment discrimination case, but – until then – this alone was generally not sufficient proof. Further evidence was needed to support the discrimination complaint. The analysis of the origin of names could be useful to show a general pattern of discrimination, but it could not prove that a specific person had been discriminated against. It is important to note that discrimination in hiring is prohibited by law in many countries, including France.

The motivation of the Court of Cassation changes everything, because the statistical analysis of the names (categorized as “extra-European” or not), is accepted as valid evidence:

having retained that the employee produced an analysis made from the single personnel register communicated by the employer for the period from March 26, 2018 to December 31, 2018 and on the organization chart of the company from which he had made statistical analyzes and had concluded that, among employees with a European surname recruited under an “interim fixed-term contract”, 18.07% had been granted a permanent contract, compared to 6.9% for employees with a non-European surname, that the employees on “interim fixed-term contracts” with a non-European surname represented 8.17% of all employees on “interim fixed-term contracts” but only 2.12% of all employees on permanent contracts for the same positions, 80.93% of employees with a European surname were under a permanent contract for only 21.43% of employees with a non-European surname, the Court of Appeal, which was not required to proceed to a research he was not asked, was able to deduce that these elements taken as a whole suggested discrimination in hiring.

Excerpt of the Court motivation, Translated by Google Translate

How reliable are statistical tools for analyzing discrimination?

A tool like NamSor makes it possible to categorize names according to country of origin with a reliability of 75%-85% on average, or even with a reliability of 95%-99% on a broader categorization such as that adopted by the Court of Cassation ( European or extra-European surname).

Other tools such as Aequitas (an opensource project initiated by the University of Chicago) allow free statistical analysis of data on gender, ethno-cultural origin or supposed “race”, or even other variables. supposed to be neutral (age, disability, marital status, etc.) The tool produces a report that demonstrates a bias, with the different metrics that confirm or invalidate the statistical value of proof.

Will the decision of the Court of Cassation set a precedent?

This decision by the Court of Cassation will certainly have an impact for all temporary work companies, starting with the largest (ADECCO, MANPOWER, RANDSTAD, etc.) and for all companies that have a high turnover in their teams, in particular consulting and outsourcing companies (KPMG, EY, PRICEWATERHOUSECOOPERS, DELOITTE, Capgemini, IBM FRANCE, Atos, Sopra Steria, Accenture, etc.) These companies have in common that they have a high turnover, and therefore significant recruitment needs. HR processes (recruitment, evaluation, promotions, dismissals, conventional termination) are highly structured as strategic processes, but also increasingly industrialized or even automated by “artificial intelligence”.

The risks of discriminatory bias exist, even when the data on gender, supposed origin etc. are not present.

The judgment of the Court of Cassation offers a possibility of proof a posteriori … if this decision were to set a precedent, it will be a sword of Damocles for all companies which have a high turnover and have not set up measurement tools on possible discrimination in hiring.

NamSor is collaborating with Entre-Autre to make patronymic analysis a robust method for evaluating human or algorithmic biases in HR processes.

NamSor is also collaborating with the French public statistics office DARES to produce statistics on discriminations in the public and private sector.

(Read in French)

About NamSor

NamSor™ Applied Onomastics is a European vendor of sociolinguistics software (NamSor sorts names). NamSor mission is to help understand international flows of money, ideas and people. We proudly support Gender Gap Grader.