Hiring discrimination and blind CVs: what the data says (2026)

It is measured and documented, not merely reported. The correspondence study conducted for DARES across around forty large companies (applications sent from April to July 2016, 1,500 tests and 3,000 applications) shows that, at strictly equivalent CV, an application with a North-African-sounding name is called back 31.5% less often than a French-sounding one: 22.8% positive replies against 33.3%. In practical terms, it takes about 1.5 times more applications to obtain the same number of replies. Of the 40 companies tested, 12 showed a statistically significant gap to the detriment of "North-African" candidates (from 15 to 35 percentage points depending on the company).

The picture has not improved. According to the 18th barometer published by the Défenseur des droits (France's equality ombudsman) with the ILO in December 2025, covering 2016-2024, people perceived as Black, Arab or North-African are 2.8 times more likely to report having experienced discrimination while job-hunting than people perceived as white — against 2.2 times in 2016, a deterioration. More broadly, 35% of respondents report unfavourable or discriminatory treatment over the past five years, and more than 9 working people out of 10 believe discrimination exists in employment.

A methodological caveat: a correspondence study measures objective behaviour (the callback rate), whereas a barometer measures perceived and self-reported discrimination. The two approaches are complementary, and they converge.

Not automatically — and the result is counter-intuitive. The only large-scale French evaluation of the anonymous CV was run through the public employment service between 2009 and 2011 by Luc Behaghel, Bruno Crépon and Thomas Le Barbanchon, covering 598 job openings and 1,268 candidates; the results were published in the American Economic Journal: Applied Economics (2015).

The paradoxical finding: making CVs anonymous widened the interview-access gap between non-minority and minority candidates, from 2.4 points to 13 percentage points. Within that scheme, anonymization actively hurt candidates of immigrant background or living in disadvantaged neighbourhoods.

The authors explain this "anonymous CV paradox" through two mechanisms, which deserve to be quoted honestly:

• Employer self-selection. Participation was voluntary: the companies that signed up were probably among the least discriminatory to begin with. Anonymity may have neutralised a form of positive discretion they were already applying.
• Loss of context. Without a name or an address, a recruiter could no longer soften a negative signal — a gap in the career path, for instance, previously read in light of the candidate's neighbourhood.

The same evaluation also qualifies the effect on gender, and precision matters here: anonymization did neutralise recruiters' tendency to favour candidates of their own sex (or age) — but since recruiters were of both sexes, those effects cancelled each other out. On aggregate, blind CVs did not improve the odds of women or older candidates. The authors concluded that "generalising the anonymous CV does not appear justified". The lesson is therefore not "blind CVs are useless", but that their effect depends on the scope (voluntary or systematic), on the type of bias targeted, and on the rest of the hiring process.

Because it was never implemented. The Equal Opportunities Act of 31 March 2006 (article 24) created article L. 1221-7 of the French Labour Code, requiring companies with more than fifty employees to review applications "under conditions preserving the anonymity" of the candidate. But the implementing decree was never issued, so the obligation never came into force.

The timeline, as recounted by the Conseil d'État (France's supreme administrative court):

• 9 July 2014 — the Conseil d'État orders the Prime Minister to issue the missing decree.
• 17 August 2015 — article 48 of the "social dialogue and employment" act amends article L. 1221-7: the blind CV becomes optional. Applications "may be reviewed under conditions preserving anonymity", rather than must be.
• 4 November 2015 (decision no. 387014) — the Conseil d'État rejects a request for a penalty payment against the State: with the obligation gone, the 2014 injunction is moot.

Today the blind CV is therefore a practice left to each employer's discretion, not a legal requirement.

Three practical lessons come out of the figures:

• Discrimination is real and persistent. It is not a matter of isolated perception: it is measured by correspondence testing (−31.5% callbacks) and confirmed by the Défenseur des droits – ILO barometer.
• Anonymization is not a silver bullet on its own. The best-documented mechanism is neutralising identifying signals at the initial screening stage. But the French evaluation is a reminder that the net effect depends on scope (voluntary or systematic) and on the rest of the process: taken in isolation, it is not enough. It is most relevant when applied systematically and paired with a structured framework — scoring grids and objective, job-related criteria.
• There is an independent data-protection argument. France's data protection authority, the CNIL, points out that in recruitment the principle of data minimisation allows collecting only data with a direct and necessary link to the job and to assessing skills. Nationality, family situation or a photograph are not used to judge someone's ability to do the job — and the CNIL has already issued a formal notice against a company for excessive collection (place of birth, nationality, family situation, past salaries).

Pseudonymizing a CV at the screening stage — removing the name, the photo, the contact details and named employers — therefore serves two goals at once: reducing exposure to bias on the most sensitive signals, and aligning the shortlisting phase with the data-minimisation principle. That is precisely what a tool like BlindCV does (genuine technical removal of the text, zero CV storage) — provided it is embedded in a structured hiring process rather than treated as an isolated gesture.