We conducted a Discrete Choice Experiment (DCE) to measure the effect of language utilized in job advertisements on the job application behavior of men and women in Latin America. Our study analyzed the probability of selecting a job to apply for when presented with a variety of sets of language-specific options.
Our study involved creating a fictitious baseline job ad in a male-dominated occupation using a male-oriented skill that is commonly used in these ads. We then compare this to job ads that use a more gender-neutral language for otherwise identical job postings. Our analysis focused on two main interventions. First, modifying the language used to describe a skill required for applicants in the job ad, or omitting it altogether, with the aim of reducing the influence of gender stereotypes. Second, utilizing gender-neutral language to signal a commitment to inclusivity, such as using gender-inclusive sentences or employing gender-neutral language in the job ad, as follows:
- Control: "Base" job ad: Male-oriented skill.
- Skill-Stereotype: T1: Skill's description -provides a description of the skill used in "Base", but does not specify the skill. T_2: No skill detailed -does not specify any skill in the job ad.
- Inclusive work environment: T_3: Diversity statement -we use a phrase that explicitly requests women and members of the disadvantaged population, but is unrelated to the skills required for the occupation. T_4: Gender-inclusive language -uses explicit language related to the feminine and masculine spelling of the skill.
Participants were exposed to nine decision sets, each consisting of pairs of hypothetical job postings that varied in terms of language. Each posting was designed to resemble those commonly found on job search engines and referred to a job within the sector and occupation that the participants had selected from a list. The job listing contained a general description of the position, information about the work schedule, holding constant the location of the job vacancy (downtown area). It did not include any other information regarding wages, the type of contract, or its duration. The reason for eliminating wages from the job postings is to avoid confounding factors related to income effects in the selection process. For this reason, we are able to measure the willingness of both men and women to apply for a job ad using a particular language, but not their willingness to pay for it.
In each decision set, participants were asked to choose between the baseline job advertisement, "Base", and one of the light-touch interventions described earlier, with no opt-out option. The aim of excluding the opt-out option was to encourage participants to make a choice between the two alternatives and provide more insights into their decision-making processes. Job seekers participants were told that the ads were fictitious. We randomly assigned two skills to each occupation chosen by the participant and created four decision sets, one for each treatment. To check for inattention, we randomly repeated one of the decision sets to evaluate if the answer was consistent. This resulted in a total of nine decision sets for each participant. In addition, we randomize by individual the position of the two job posts in each screen (left/right), and the order of appearance of all nine screens.
This design allows us to elicit revealed preferences for job ads that use different types of language, instead of measuring stated preferences of working-age job seekers in the five Latin American countries.