The wording in job adverts can discourage certain segments of the population, but here’s how to de-bias them
Are you looking to recruit a ‘dynamic leader’ or a ‘committed people person’? Chances are you’re just looking for the best person for the job. But the choice of language used in the job description could be alienating and dissuading the best – and most diverse – candidates from even applying.
Recent research from Adzuna revealed that 60% of businesses showed significant male bias in the wording of their job adverts. This research was based on a study by academics Gaucher, Friesen and Kay, which found that job descriptions with more masculine wording were less likely to appeal to female applicants. It wasn’t for themost part that female candidates assumed they weren’t up to the job, the research found. Rather they – consciously or unconsciously – were less likely to feel they’d belong at such an employer, and didn’t want to work for a company whose first impression was one of being biased in favour of men.
And so debate on the issue is hotting up. The UK government recently announced a trial of gender-neutral language to define science, technology, engineering and maths apprenticeships to encourage more women to apply. A pilot will apply gender-neutral language to 12 apprenticeship standards.
But while most HR leaders are aware that biased language exists in job descriptions, many don’t know how to fix this. Part of the problem is an inability to identify biased language because of its subtlety. Words that seem innocuous are often rooted in societal conditioning.
A 2017 analysis of 77,000 UK job adverts by Totaljobs revealed ‘lead’ to be the most common male-gendered word used in job specs, while ‘support’ was the most used female-gendered word. According to Gaucher, Friesen and Kay, popular recruiting adjectives such as ‘ambitious, assertive, decisive, determined and self-reliant’ are male gendered. While words like ‘committed, connect, interpersonal, responsible and yield’ are considered female gendered. For instance, in a male-gendered job description a company might be described as ‘a dominant engineering firm that boasts many clients’. Whereas the female-gendered version could read ‘we are a community of engineers who have effective relationships with many satisfied clients’.
So how can HR de-bias a job description to make the language gender neutral? According to Andrea Singh, HR director of BAM, the first step is to focus on gender-coded words. Job titles should be neutral and descriptive language should give equal weighting to male- and female-coded descriptors, she explains. However, Singh also points out that de-biasing a job description goes beyond replacing adjectives. Employers need to make sure that the requirements listed are actually necessary, because “women will typically only put themselves forward for a job when they meet 100% of the criteria”.
But with unconscious bias ever present there are questions around whether it’s possible for humans to conduct this de-biasing. Singh believes that with the right training it is. But she admits the best results come when software and learning are combined. “Technology brings information and suggestions to the fingertips but job specs need to feel authentic. The people writing and editing specs need to be trained to spot the bias too,” she says.
However, Richard Marr, co-founder and chief technology officer of Applied, doubts whether training a person to remove biased language can be as effective as relying on dedicated software. “The evidence is pretty weak that training is effective,” states Marr. “Processes trump training and tools trump processes. With training you’re just expecting people to do the right thing.”
That said, the trouble with using software is that neither Applied nor its competitors AdPro and Textio currently extend their job description analysis beyond gender to include other demographics such as BAME, LGBTQ+, disabled or economically-disadvantaged candidates. Applied is working with Google to expand its analysis tool to incorporate ethnicity (and other dimensions). But until such tech is available removing gendered language from job descriptions can still have a positive impact on other diverse groups, Singh believes.
“I think language can be looked at in the same way. Masculine phrasing might also be off-putting for candidates from particular BAME backgrounds where their culture doesn’t typically fit with this type of approach,” she says.
It’s a view shared by Marr. He explains that a job analysis tool will also assess the readability and density of a job description, scoring it for how many syllables, words and sentences it contains. His thinking is that the more readable the job spec, the more inclusive it is likely to be.
“There are heavy socio-economic correlations,” notes Marr. “If you look at people who have low incomes they will have less access to desktop computers and are more likely to rely on their phones and to live in a distracting environment. Each of those things adds a cumulative layer that results in something quite substantial.”
So there are certainly steps that can be taken. But, in an age in which many urge the need to move away from binary definitions of men and women, is so-called male and female language really meaningful anymore? Or is it just another theory to get bogged down by?
Adrian Love, recruitment director for the UK and Ireland at Accenture, certainly feels male and female language is still a ‘thing’. He points to Accenture figures showing an increase in female job applicants from 34% to 50% since 2014, thanks in part to the de-biasing of job specs.
“The impact has been very positive. But there are no silver bullets here. It has to be part of a wider inclusion and diversity programme,” he says.
It’s a similar story from Applied, with Marr reporting that the tool has helped trigger an estimated 10% to 15% swing towards female candidates. Singh also reports a significant increase in female applicants since implementing de-biasing.
“This shows that [using] gender-neutral language is affecting the talent we can attract,” she says, adding that de-biasing could now be taken further. “We now need to delve into the data in more detail… and analyse the next stages in the process to see if we have more women being shortlisted, interviewed and ultimately selected.”
After all a gender-neutral job description can only go so far if, when a candidate is successful or unsuccessful in their application, the language in the feedback or job offer sees a return to bias.
Both Singh and Love concede that their job description writing tools are unable to analyse interview feedback. But this is where training comes into play, they say.
“Software raises awareness and can point out bias that people may miss,” says Singh, but it’s also important teams are trained to spot it elsewhere in recruitment materials.
Love agrees: “[It’s] not just about one action, it’s about looking at every element throughout the recruitment process. There are opportunities to drive inclusivity end to end, but job descriptions are important because they’re a gateway for candidates.”
Analysing bias in the Bank of England governor job advert
Later this year Bank of England governor Mark Carney will stand down. He’s the 120th white man out of 120 individuals to have ever filled the role, and so the institution has been heavily criticised for embodying a ‘stale, male and pale’ image of finance. By its own admission it will fail to meet any of its diversity targets this year. So with calls to appoint a female to the position for the first time is the language in the role’s job description gender biased?
Not according to Applied’s job description analysis tool. Following the appointment of diversity specialists to head up the search for Carney’s replacement, HR magazine analysed the job description to see if the bank’s commitment to diversity extends to its recruitment materials. It scored a respectable 84% for inclusivity and contained an equal amount of male-gendered and female-gendered words.
Marr says that language falls into two categories: agentic and communal. Agentic language is considered male coded. In this advert agentic traits found were words like ‘confidence, decision, lead and determination’. The communal traits were female-coded words such as ‘responsibility, commit, communicate, and understanding’.
Marr argues that performance evaluation and leadership development should also be defined in a way that balances both sets of traits. “Companies often define success for leaders along agentic lines and measure performance and promotion that way, even though communal traits are just as valuable in leaders,” he says.