Better, fairer recruitment with behavioural science.

Behavioural science can help create a better, fairer workplace, but we need to #ThinkSmall to deliver on its potential. There are few places where that’s truer than in recruitment, so we're sharing some research and insight you can use to re-think each stage of the process. We hope it will help you take few steps towards a more inclusive workplace.

 

Attraction Strategy

BEHAVIOURAL INSIGHT: The messages you use to promote your company may appeal to some demographic groups over others. This could impact the inclusion of your job-specific advertising, as well as the generic branding on your corporate website, social channels and recruitment collateral. 

RESEARCH EXAMPLE: The Behavioural Insights Team explored the effectiveness of four common career motives with the Chattanooga Police Department. These messages were delivered as postcards but they could easily be social media assets or headlines on a website. As you can see below, there was a significant difference between the response to each message. By exploring that effect the team succeeded in attracting more BME candidates (underrepresented in the dept. at the time of the research)

Postcard 1: Service

Postcard 1: Service

Postcard 2: Impact

Postcard 2: Impact

Postcard 3: Challenge

Postcard 3: Challenge

Postcard 4: Career

Postcard 4: Career


 
bst_03Artboard 1@2x.png
 

TAKEAWAY: Why should people work for you? You can't focus on everything, particularly in a short, generic message like the above. The problem is when the lead motive you use to represent your brand is particularly popular with a demographic that already dominates your organisation (a situation not unlikely if it was generated from internal research). When you use that motive to grow your team you may strengthen your over-representation. 

The answer is to experiment with your communications and explore their impact by demographic. The advantage is that it will help you become more inclusive AND save you money on your media spend by improving your conversion rates (the Behavioural Insights Team tripled applications to the Chattanooga Police Department as well as increasing the diversity of the candidates). This type of experimentation is more accessible than ever in the digital age and is already being deployed as standard practice in consumer communications. Many social media platforms (Facebook, Instagram etc.) even have A/B testing built into their platforms. You may still need some support with randomisation and experimental design, but these channels offer the best opportunity for your first exploratory steps...

 

Job Descriptions

BEHAVIOURAL INSIGHT: The language you use in your job descriptions may have a significant influence on who applies to that role and will sometimes systematically discourage applications from underrepresented groups.

RESEARCH EXAMPLE: This area has several well-established solutions, most notably from Textio, who deploy a massive dataset and machine learning to help you write job descriptions more inclusively in real time (with a particular focus on gender). You can see how organisations can inadvertently betray negative cultural attributes in the research below. Without tracking this language you’ll never know the unseen effect of your words.  

textio.png

TAKEAWAY: You can run a few job descriptions through the Textio system for free to get an idea of its functionality and potential, but you can also use a more scientific approach to explore other areas of diversity and inclusion. We would suggest you define a specific problem first (i.e. explore your organisational data to identify where you are struggling to attract underrepresented groups in attraction phase), before addressing how and why your job descriptions might be holding you back. 

 

Application and CV Screening

BEHAVIOURAL INSIGHT: Subtle information signalling gender, ethnicity, faith and social class impacts the likelihood you will be given an opportunity to show your strengths at assessment stage. When you accept CV’s with data including name, hobbies, contact information, head-shot and academic institution, you leave yourself vulnerable to the influence of explicit and implicit bias in the screening process

RESEARCH EXAMPLE 1: 1500 versions of the same CV were sent to organisations in Germany by Doris Weichselbaumer at the University of Lintz. The only difference was the photo and the name: 

Recruitment Discrimination Copy.png

RESEARCH EXAMPLE 2: In a similar study by Lauren Rivera and Andras Tilcsik, cues were changed to signify gender (via name) and social class (via name and hobbies). Once again, although the CVs were otherwise exactly the same, both men and people from higher class backgrounds were significantly more likely to receive callbacks from employers. 

TAKEAWAY: The response to this type of research requires judgement and pragmatism. We support the overall challenge to the value of the traditional CV, and many emerging tech providers make the transition away from that format more accessible than ever (see Applied for our favourite example). But we also acknowledge that some people may require a staged approach with existing suppliers. Either way, we would encourage you to take steps to remove information that is not predictive of performance in your screening process. As above, you may not realise the damage it brings to the inclusivity of your hiring process. 

 

Shortlisting 

BEHAVIOURAL INSIGHT: Many organisations will conduct assessments to 'narrow down' a shortlist for final interview, deploying things like personality questionnaires, verbal and numerical reasoning, cognitive tests and situational judgement tests (SJT). We can't cover all of these in detail, but consider that under-represented candidates may be disadvantaged in all these assessments by a threat of stereotyping. Just imagine you are in an environment in which no one looks or sounds like you and you are being evaluated for your potential to join that group — how much would you begin to doubt your instincts?

RESEARCH EXAMPLE: We'll return to the Behavioural Insights Team and another Police Service (the Avon and Somerset Constabulary) for an extraordinary example. Drawing on previous empirical evidence, the researchers identified that 'stereotype threat' might be causing a low SJT pass-rate among BME applicants. They designed an experiment to test whether this could be overcome by 'positive priming' - encouraging candidates to reflect what might make them a good addition to the force before the assessment. 

The below is the performance difference between BME and Non-BME applicants who received 'positive priming' in their SJT invite, compared to those that didn't. The effect on the performance on BME candidates was dramatic: 

 
Situational Judgement Test — pass rate without intervention.

Situational Judgement Test — pass rate without intervention.

Intervention - two additional questions in the email invitation.

Intervention - two additional questions in the email invitation.

Situational Judgement Test - pass rate with intervention.

Situational Judgement Test - pass rate with intervention.

 

TAKEAWAY: We're wary of oversimplification in this section. The above results are astounding and indicate the unseen disadvantages faced by some over others in the assessment process. But there are also no general answers. We know enough about The Behavioural Insights Team to assume they were able to identify this intervention by approaching the challenge set by the Avon and Somerset Constabulary with an open, inquiring, scientific mind. This means focusing on a clear challenge and research question, gathering evidence from the organisation and its employees, and drawing parallels with literature from behavioural science before taking action. 

 

Final Interviews  

BEHAVIOURAL INSIGHT: Consider that the diversity of your interview shortlist may have a disadvantaging influence on underrepresented groups.

RESEARCH EXAMPLE: Johnson et al (2016) found that the proportion of under-represented candidates in an application process has a significant influence on who is successful, beyond the increase you would expect due to probability.  They studied the hiring decisions of 598 job finalists across a three year period, of whom 174 had received an offer: 

  • The odds of hiring a woman were 79.14 times greater if there were at least two women in the finalist pool.

  • The odds of hiring a minority candidate were 193.72 times greater if there were at least two minority candidates in the finalist pool.

This extreme effect held no matter the size of the pool (six finalists, eight finalists, etc.), and these analyses excluded all cases in which there were no women or minority applicants. However, this research was never published outside the Harvard Business Review, so we need to be cautious and critical in it’s evaluation.

TAKEAWAY: Create balanced shortlists in areas where you are underrepresented. everything you can to limit bias in the interview process. This might include structured questions, multiple interviewers conducting separate interviews, interviewers providing feedback individually before discussing as a group, and reviewing candidates in comparison to each-other rather than an 'imagined ideal'. Finally, consider quotas in your shortlisting process to redress unequal opportunity and assessment.

 

While all these stages need separate attention there are things that unite them. Define the problem before you jump to action, gather evidence from employees, organisational data and academic literature, and test your assumptions with the power of experimentation. We bring this #ThinkSmall and #TestLearnAdapt approach to everything we do at MoreThanNow.

We're here to help if you need us.