Welcome to the show everybody! Today we're stepping into a topic that gets argued about loudly: diversity in hiring. Some people say these programs lower the bar. Others say they're the only way to widen it. And I want to start with a study that I think should make both sides stop and look at the data. David Larcker at Stanford and colleagues from Chicago, Yale and the University of Washington looked at nearly thirteen hundred diversity related controversies at three hundred fifteen U.S. public companies between 2008 and 2022.
And what happened after the controversies?
The talk went up. After a controversy, companies were nearly eighteen percent more likely to mention diversity in their proxy statements. But the hiring barely moved. Hiring of women and people of color rose by less than one percentage point over the next two to three years, mostly in junior and back office roles. Senior hiring actually fell. And after netting out departures, overall workforce diversity rose by just four tenths of a point.
Four tenths of a point. That tells you everything about the difference between messaging and measurement. Messaging is what a company says. Measurement is what its hiring process actually does, decision by decision. And whichever side of the debate you're on, you should want the second one to be honest.
So if the statements aren't working, what does?
Most organizations answer that question with training. Send the hiring managers to a workshop on unconscious bias and hope they come back different. The research is fairly clear that this is the weaker lever.
There's a 2025 systematic review on exactly this. Isabela Merla, Fiona Gabbert and Adrian Scott at Goldsmiths, University of London looked at thirty eight studies of interventions to reduce bias in high stakes professional judgments, most of them in workplace settings. They split the interventions into two kinds. Individual level ones, which try to change people's attitudes or awareness. And systemic ones, which change the decision process itself, things like decision protocols, standardized scoring rubrics, or changing how information is presented to the person deciding.
And the systemic ones won.
Clearly. Fourteen of the eighteen systemic studies showed strong effects. That's about seventy eight percent. For the individual level approaches, it was seven out of eighteen, about thirty nine percent. So changing the process worked about twice as often as trying to change the person.
That matches everything I've seen in organizations. You can't train a reflex out of a tired hiring manager at four o'clock on the sixth interview of the day. What you can do is remove the opportunity for the reflex to decide. Ask every candidate the same questions. Decide in advance what a strong answer looks like. Score each answer before anyone discusses the candidate. Set the weighting before you see a single application. None of that requires anyone to become a better person. It just requires a better process.
There's a small but striking example from Yale. Researchers looked at applicants to a pediatric emergency medicine fellowship. The program's standard scoring tool, which leaned on things like test scores and academic honors, showed a significant difference between racial groups. But the structured, behavior based interview scores didn't show a significant difference at all. It's one program and a small sample, so I wouldn't overstate it. But it points in the same direction.
It does. And notice what that example says about the old argument that fairness and merit are in tension. A structured process isn't a lower bar. It's a clearer bar. When every hire can show exactly why they were chosen, against criteria written down before anyone applied, the label people throw around, the idea that someone was a diversity hire rather than the best candidate, loses its footing. Structure protects the candidate and the decision at the same time.
We covered the legal side of this in a recent episode, so I don't want to repeat that. Let's talk about measurement instead. If a company wants to know whether its process is actually fair, what should it be counting?
Not the things that are easy to fake. The size of a diverse applicant pool is easy to inflate. A statement is easy to publish. What's hard to fake is the pass through rate at each stage of the funnel, broken down by group. Who gets from application to screen, from screen to interview, from interview to offer. If one group drops off sharply at one particular stage, that stage is where your process is making a decision you didn't design.
And after the hire, too. The Stanford study found turnover increased among women and people of color after those controversies. So the people were hired, and then they left.
Which is why early turnover by group belongs on the same dashboard. If people are hired and gone within the first year, the process didn't succeed. It just moved the problem down the hall.
What about the very top of the funnel? The job description itself?
That's where a lot of quiet filtering happens. Degree requirements the job doesn't actually need. Vague phrases like culture fit that let anyone justify anything. Every requirement should trace back to what the role actually demands. If it doesn't, it isn't measuring merit. It's measuring familiarity.
And this is where behavioral data comes in, as long as it's the right kind.
As long as it's validated and job related, yes. Every candidate is measured the same way, against what the role demands, not against who the interviewer happens to relate to. Ask any vendor for the validation and adverse impact evidence in writing. But I want to be clear. It's one structured input. It complements skills and experience. It never replaces them, and it should never be the sole basis for a hire.
So the honest summary is: stop trying to change people's minds in a workshop, and start designing a process that can show its work.
That's exactly it. A fair process isn't one that says the right things. It's one that can prove what it did.
If you want to see what a validated, job related input looks like, start with the person you trust most on your team. OAD's free Top Performer Profile takes seven minutes, results come back in twenty four hours, and there's no credit card. That's O-A-D dot A-I.
Good intentions write the statement. A good process writes the record.
Perfect place to leave it. Thanks for the wisdom.
Always a pleasure.