Can job ad language help explain the gender gap in the Indian labour market?

Women applying for jobs tend to earn a lot less than men of the same age and education qualifications. Understanding the factors contributing to this disparity is crucial for addressing gender inequality in the labour market. This article explores one such factor: the wording of job advertisements and how it is linked to applicant behaviour. It finds that women might be deterred from applying to high-salary jobs due to implicit gender associations, together with explicit preferences.

This is the fourth post of a five-part series to mark International Women’s Day 2025.

Why do women apply to lower-salary jobs compared to men of same age and qualifications? This question is important, as recent research has shown that differences in applied-for jobs explain a substantial portion of the gender wage gap after accounting for individual-level observables in multiple labour markets.¹ Our study (Chaturvedi et al. 2024) builds on this by analysing job ad language in India, where explicit gender preferences are still prevalent despite legal frameworks discouraging discrimination. Studies in Western contexts suggest that masculine-coded words such as competitive and ambitious tend to deter female applicants, whereas feminine words like supportive and compassionate attract them.² However, little research has systematically examined what attributes employers associate with women and men or how they are linked with the application behaviour of candidates.

Data and methodology

We analyse data from 157,888 job ads posted between July 2018 and February 2020 on an Indian job portal, along with 6.45 million applications submitted to these ads. The job postings in our dataset are for relatively high-skill positions, with a mean (average) posted wage that is 21% higher than the average earnings of a nationally representative and comparable sample of workers in urban India from the Periodic Labour Force Survey (PLFS), 2017-18. Job-seekers on this platform are younger (with an average age of 24 compared to 36 in PLFS) and more educated (86% have a graduate or higher degree compared to 32% in PLFS).

To classify gender preferences of employers, we categorise job ads into three groups: those that explicitly state a preference for female candidates, for male candidates, and those without explicitly stated gender preferences. Next, we use a ‘logistic regression classifier’ on text in job ads to construct measures indicating whether this text is predictive of a female or male preference. We refer to these as a job ad’s implicit femaleness or maleness, which can be estimated even for job ads without explicitly stated gender preferences. We then measure how these are linked to the gender composition of applicants and posted wages.

Key findings

We find that 7.7% of job ads explicitly state a gender preference, with 4.2% seeking female candidates and 3.5% preferring men. We find that implicit femaleness is high in job ads with stereotypically female job titles such as beautician, personal secretary, and schoolteacher. At the same time, implicit maleness is high in job ads with stereotypically male titles such as cargo loader, delivery executive, and network engineer. Given that we use the entire job ad text to construct these measures, implicit femaleness and maleness can vary for jobs with the same title but a different job description. For instance, Figure 1 shows two explicitly gendered job ads in our data that have the title of ‘business development manager’: we find that implicit femaleness is high for the job ad emphasising appearance or communication skills, while implicit maleness is high when the job involves fieldwork.

Figure 1. Job postings with an explicit gender preference

Notes: (i) Panels (a) and (b) show job postings with an explicit female and male preference, respectively. (ii) Words highlighted in red reflect female associations while those in blue reflect male associations. (iii) Colour intensity reflects the strength of the attached gender association, with darker shades indicating a stronger association.

Despite a higher fraction of jobs requesting women, such jobs offer lower salaries than those requesting men or jobs that do not specify a preference. We find that women on this portal apply for jobs with 3.5% lower posted wages, on average, than men of the same age and education qualifications, who are located in the same state. A substantial portion of this gap (approximately 45%) can be explained by women applying to jobs in low-salary occupations and low-salary jobs in specific parts of the country, but the rest remains unexplained. A possibility that we investigate in our work is that women might be deterred from applying to high-salary jobs if such jobs explicitly state that men are preferred for the role or if the job ad wording indicates this even without including an explicit request for men. We estimate that explicit gender requests account for an additional 7% of this gap, while implicit gender associations, together with explicit preferences, explain 17% of the gap, after accounting for a range of controls.

We further investigate what kind of words employers associate with men and women and how they contribute to the gender wage gap in applications.³ We find that hard skills predictive of a female preference such as those associated with beautician roles, accounting tasks and software (ledger, expense statements, Tally, Zoho), computer-related design and communication tools (MS Office, Corel, AutoCAD), and keyword analyses are linked to lower wages but attract more female applicants. On the other hand, job postings that mention male-associated flexibility constraints, such as night shifts, weekend work, relocation, and travel requirements, offer higher wages but receive fewer applications from women.

Similarly, soft skills and personality-related words in job ads show a gendered pattern as jobs requesting women focus on communication, coordination, and interpersonal skills while those requesting men highlight assertiveness or leadership such as liaisoning, negotiating, supervising, and motivating. Jobs requesting women emphasise politeness, patience, adaptability, and punctuality and include physical traits such as height or a nice smile. Those requesting men include personality traits such as being energetic, enthusiastic, resilient, passionate, resourceful, prompt, honest, and methodical and physical traits such as chest measurement. We find that the presence of soft skills and personality-related words associated with women is not associated with a higher female share in the pool of applicants to a job ad. This provides suggestive evidence that employers' gender stereotypes (as captured by gendered words) may be based on distorted or inaccurate beliefs – at least regarding some characteristics of an ideal female versus male job applicant.

Policy implications

Recent studies by Card et al. (2024) and Kuhn and Shen (2023) provide causal evidence that removing explicit gender requests can increase gender diversity in the workplace without negatively affecting firm survival, employment, or wages. However, our findings suggest that simply banning explicit gender requests in job ads may not be enough – employers may still use implicit signals, such as gendered words, to shape their applicant pools. While regulatory interventions such as bans on explicit gender requests are a step forward, they may need to be accompanied by broader efforts to challenge gendered job perceptions among employers and recruiters. Public awareness campaigns or recruiter training programmes could help reduce gender-stereotyped language in job ads and encourage more inclusive hiring practices.

Our research examines young Indian workers entering the labour market soon after completing their university degree. Several papers (Kahn 2010, Oreopoulos et al. 2012, Rothstein 2023) document the persistent effects of initial labour market conditions (such as a recession) when young workers enter the labour market on long-term economic outcomes. This suggests that the gender differences at a relatively early career stage we study are also likely to have important implications for lifetime earnings and financial independence of women.

Notes:

Analysing Danish administrative data, Fluchtmann et al. (2024) find that differences in the jobs to which individuals apply, explain 79% of the residual gender wage gap in typical earnings and 73% of the residual gap in realised starting wages – after controlling for individuals' observable characteristics.
Gaucher et al. (2011) show variants of hypothetical job ads to students to examine the role played by gendered wording, based on character traits stereotypically associated with women and men, in maintaining gender segregation.
To identify which words in job ads contribute to the decisions of the logistic regression classifier (or gendered words), we use the Local Interpretable Model-agnostic Explanations (LIME) algorithm proposed by Ribeiro et al. (2016).