Several concurrent hypotheses have been proposed in order to take into account the persistence of under-employment of persons aged over 50, in France. In this paper, we are interested in the labour demand. We construct and implement an experimental protocol allowing testing four hypotheses currently discussed: the unobservable distance from retirement by the recruiter; the supposed depreciation of skills in a context with repeated technological shocks; the supposed inability of professional retraining, eventually differentiated by gender; a taste discrimination against older workers, as the result of social norms. To test each of these hypotheses, we have conducted four separate campaigns of testing for evaluating them. Between mid-January and mid-August 2015, we have sent, 6 361 applications allocated on seven occupations and twenty-eight profiles. The research investigates a statistical processing of results of these testings. Our results indicate which must necessarily combine several hypotheses to take into account the under-employment of seniors, in France: the difficulties of professional retraining following a shock in the employment career; the social norms inducing a negative representation of the older workers through the employers. Seniors are both victims of the depreciation of the human capital and age discriminations in access to employment, which will especially penalise men in their professional retraining. The classical explanation by the distance from retirement is the only one which isn’t statistical validated.