Words Matter: Gender, Jobs and Applicant Behavior

Using Domain-Specific Word Embeddings to Examine the Demand for Skills

“…On the other hand, working in factories and warehouses, mechanical and electrical engineering, quality inspection, machine learning, and knowledge of programming languages is associated more with men by employers, as reflected in the word embeddings. This is consistent with the findings in Chaturvedi et al (2021) who uncover words predictive of explicit gender requests in job ads by employers using data from a di↵erent job portal and employing a di↵erent method.…”

“…8 Word embeddings can also be used to capture gender associations in text corpora that might reflect cultural stereotypes (Bolukbasi et al, 2016;Caliskan et al, 2017;Ash et al, forthcoming). Chaturvedi et al (2021) show that more women apply to job ads when the ad uses words predictive of a female preference by the employer. Several other papers also provide evidence that women's application decisions are a↵ected by the wording in job ads (Abraham & Stein, 2020;Kuhn et al, 2020).…”

Stereotypes of older workers and perceived ageism in job ads: evidence from an experiment

“…Gaucher et al (2011) found that job ads for male-dominated occupations used words associated with male stereotypes (such as 'leader', 'competitive', or 'dominant') more frequently than advertisements for female-dominated occupations, and women found job advertisements less appealing when they contained more masculine than feminine wording (Bem and Bem, 1973;Gaucher et al, 2011). Chaturvedi et al (2021) use machine learning to study job ads, identifying words that are predictive of a gender Though they did not focus on job ads, Hanson et al (2011) and Hanson et al (2016) study language used by mortgage originators and connect this language to their behavior. Hanson et al (2011) study subtle discrimination through 'keywords' used by landlords responding to prospective tenants.…”