Building Diverse Teams: Stop Saying Its a Sourcing Problem

There’s a lot of discussion these days about the problems with how we hire and the impact that’s having on diversity. 

And rightly so. 

In the UK alone, just 6% of FTSE 100 bosses are women, and more are called Steve than come from an ethnic minority background. 

But this isn’t just a problem seen in the private sector, and it’s not just race and gender either. A recent study by the Sutton Trust and the Social Mobility Foundation found that while less than 1% of the UK population studied at Oxford or Cambridge universities, but their graduates make up a staggering 71% of senior judges, 57% of the Cabinet, 56% of heads of departments, and 36% of the top jobs in the media. 

As a talent lead, a CEO,  or even a concerned hiring manager - a question on your mind is probably: “where are we going wrong?” 

‍

The attractiveness of attraction

Many people believe that their failed attempts at hiring for diversity are caused by a problem with attracting candidates from a wider variety of backgrounds. One of ‘sourcing’, to use the HR lingo. 

Probably every week someone comes to us saying “We’re not biased in who we select once they’re in the process, but we just can’t get [insert under-represented group here] to apply in the first place!”.  

It’s a seductive explanation because it suggests the issue is at least 50% the candidate’s problem, and it avoids the awkward thought that something is going wrong internally.

While it’s absolutely the case that how you craft a job description, how you design your employee value proposition, and your approach to candidate attraction will impact on who chooses to apply - if you’re using a conventional CV sift and interview process, there’s a very strong chance your selection process is actually causing most of your problems.  Here’s why.

‍

Discrimination in hiring: the search for answers

For decades, researchers have been trying to understand how level the playing field is for jobs. 

Not only because it goes directly to the question of fairness, but also because your job opportunities have a large and persistent impact on your psychological and economic prospects. And these filter across generations, ultimately driving social (im)mobility. 

There are countless inequities in educational and economic opportunities that cut across gender, race and other demographics. For example, as detailed by the OECD, disadvantaged children are far more likely to go to poorer and less well-resourced schools, which in turn affects their skill development, and consequently, their job prospects. These structural inequalities are real and powerful.

And studies have shown that you can put off candidates from applying to jobs by using gendered language and crafting your job description in ways that imply a certain type of applicant over another. (This is why we created our job description analysis tool).

But academics have sought to understand whether discrimination exists, even if you have two completely equally qualified candidates (i.e. you could level the prior playing field and interest).

‍

A simple experiment

To study this, researchers designed a remarkably simple experiment: create fictitious candidate applications, submit them to real employers, and then measure who gets through to the next stage (usually an interview). 

In its basic form, the fictitious candidates are exactly equal in their educational backgrounds, work experience, and other qualifications; but differ in some key characteristic - for example, ethnicity (as indicated by name).  

By making all candidates of equal quality, they’re then able to say with confidence that any differences they see in the interview rate is down to that key characteristic. And because they’re aggregating over many employers and randomising the applications, they’re also able to rule out that the difference was down to the idiosyncrasies of a specific employer. 

A notable example of this type of study was done in Germany in 2016. Researchers wanted to understand not only the impact of ethnic prejudice but also how that intersects with religious discrimination (effectively getting at intersectional inequality). In their study, they created three candidate profiles - all exactly equal - and randomly sent applications out to over 1000 employers. Since pictures are commonplace on applications in Germany, they could use the candidate’s name and their picture to understand discrimination. 

First, they varied the ethnic association of the candidate’s name, comparing a German-sounding name (Sandra Bauer) with a Turkish-sounding name (Meryem Öztürk).  They found that on average, Sandra was offered an interview almost 20% of the time, but the exact same application from Meryem only got an interview 14% of the time. 

They also evaluated whether giving different signals about the candidate’s religious background would have an impact. The results were stark. When Meryem was wearing a headscarf (indicating she was Muslim), her call-back rate dropped to just 4%. 

Or, put another way, an equally qualified candidate with an ethnic minority surname wearing a headscarf would have to send out 4.5 times more applications to get the same chance of securing an interview.

But the researchers didn’t stop there. 

They were also keen to understand if these effects differed by job, seniority, company type and location. Sadly, statistically significant discrimination persisted across all six cities they studied closely, and across German-based and international organisations. 

The bias was also worse for more senior jobs. They sent applications for junior accountant and chief accountant roles, and found that the degree of discrimination against headscarf-wearing Meryem almost doubled. In a senior role, she needed to send closer to 8 times as many applications to get the same acceptance rate as Sandra. 

Other researchers have played with different variables to try to unearth the source of this discrimination. In Canada, for example, academics have tested whether discrimination against candidates with foreign-sounding names is a result of employers assuming that a foreign-sounding candidate struggled with English. So they tested if rates were different depending on whether they were educated in Canada or not (and whether they’re applying to higher or lower skilled jobs, and to larger or smaller organisations). 

They found that having an Asian-sounding name but a Canadian education resulted in a 28% drop in your likelihood of getting an interview. This grew to a shocking 63% drop if you had an Asian-sounding name and were educated abroad. 

As in the previous study, they tended to find worse discrimination in “higher skilled” jobs (accountants, civil engineers, and sales and marketing managers).

But they also tested organisation size. They found that discrimination is pervasive across all organisation sizes, but is up to twice as acute in small and medium sized organisations than in large organisations. This matters because around 70% of people work for small or medium-sized organisations. 

Possibly the most concerning finding in their study is that discrimination can’t even be beat with higher qualifications. 

In larger organisations, giving the Asian-named applicant a master’s degree only brought them just about even with their Anglo-named counterpart without the degree; in smaller organisations, even that isn’t enough to level their chances. This mimicked a US study which showed that a candidate with an African American-sounding name needed 8 more years of experience to reach the same interview rate as a White-sounding counterpart.

Simply being committed to diversity is no panacea either. Another US study has shown that Black and Asian candidates who ‘Whiten’ their resumes (that is, adopt Anglo-Saxon-sounding names and/or remove ethnic signifiers from their extra-curricular work experience) have more success applying to jobs, even in organisations with strong commitments to diversity. In fact, they found just as much discrimination in organisations which contained an explicit pro-diversity message in their job advertisements.