Did you read the whole paper? They control for the type of job in the multiple studies conducted in this paper - the job same job is advertised with differently worded postings. They also ran tests with male-dominated professions, female-domainated professions, and neutral professions - the result was the same.