DEI Is Spending Billions and Moving Millimetres

McKinsey's "Diversity Wins" series - the most-cited source for the business case for diversity - has failed independent replication. Jeremiah Green and John Hand, working across S&P 500 firms using McKinsey's own diversity measures, found no statistically significant relationship between executive diversity and financial performance across six different performance metrics. They published their findings in 2024. Oddly, the response from organisations has been to keep citing the original McKinsey reports. 

This is a red flag about mechanisms for change. The research that has failed to replicate is the one that treats demographic diversity as the variable - the idea that adding people who look different to a leadership team will, by itself, improve financial returns. Robin Ely and David Thomas have been saying for years that there is no rigorous evidence for that claim. What the evidence does support is something more specific, and more demanding: diverse teams outperform when the conditions that allow difference to translate into better thinking are actually present. Those conditions are inclusion, flexibility and how power is held, used and shared. They are the variables most DEI programs have barely touched. 

Decades of investment

The conventional DEI toolkit runs on three interventions: awareness or bias or leadership training, demographic representation targets, and compliance activity. Two of these have no credible evidence base linking them to performance improvement. The third - representation in senior leadership specifically - does. The BCEC and WGEA's 2020 Gender Equity Insights research, using six years of ASX-listed company data and instrumental variable methodology to establish causal direction, found that increasing the share of women in top-tier leadership increased the likelihood of companies outperforming on multiple profitability metrics, and that appointing a female CEO produced on average a 5% increase in market value - approximately AUD$79.6 million. The 2025 update found gender-balanced leadership linked to company value gains of close to $93 million. This is the Australian evidence base that distinguishes representation in senior leadership from the broader DEI activity that surrounds it - and it points precisely to where the performance mechanism actually sits: in the rooms where decisions are made, not in training programs designed to change what people know.

On demographic targets, the Larcker et al. study (Stanford, 2025) examined corporate responses to DEI controversies and found firms dramatically increased DEI language - in proxy statements, CSR reports, social media - while workforce composition rose by less than one percentage point, concentrated in junior and non-core roles, with diverse departures also rising. The researchers called it “diversity washing”. The organisations that avoided the sustained stock underperformance seen in their peers were the ones making substantive structural changes - not the ones talking more. The correlation between DEI rhetoric and DEI outcome is essentially zero. The correlation between structural accountability and outcome is measurable.

What the conventional model has produced, reliably, is activity-without-outcome: high completion rates, high spend, minimal movement in the numbers that matter. Gartner found that only 12% of HR leaders believe their organisation has been effective at increasing diversity representation. This, after decades of investment, programs and strategies. The diagnosis the evidence supports is not that organisations haven't tried hard enough. It's that they've been trying the wrong things. 

Connecting Diversity to Performance

Ely and Thomas's synthesis cuts through the business case debate precisely. Diversity yields performance benefits only when organisations create a climate where people can express themselves, where bias is actively countered, and where the genuine differences in perspective and knowledge that diverse teams bring are used to do the core work better. They call it the learning-and-effectiveness paradigm. Their finding: organisations that simply add diverse people to existing structures and assume performance will follow are not activating this mechanism. They are creating a condition where difference is visible but not heard - which is neither inclusion nor performance improvement.

The psychological safety research is another evidenced part of this argument. Amy Edmondson's foundational work, tracking 51 teams and followed by a meta-analysis of hundreds of studies, established that psychological safety - the shared belief that the team is safe for interpersonal risk-taking - drives learning behaviour in teams, which in turn drives team performance. Google's Project Aristotle analysed 180 teams across 35 statistical models and ranked psychological safety as the single strongest predictor of team effectiveness - above structure, clarity, meaning and individual talent combined. The finding is that safe teams make better decisions, surface more problems, and generate more ideas, because the people with the relevant knowledge are actually able to contribute it.

Here is the inclusion implication that most DEI programs have missed. Psychological safety is created by how power is exercised in the room - by whether the person with the most authority invites challenge, responds well to bad news, and makes it genuinely safe for people to disagree. That's a leadership behaviour, and it's determined by how conscious leaders are about the power they hold and how they use it. The organisations that have built genuine psychological safety have done it by changing how leaders behave in specific moments, not by changing what people know about unconscious bias. 

Flexibility and Power

Giving people genuine control over how and when they work - genuine control over conditions, with outcomes as the measure rather than presence - consistently produces better output, higher discretionary effort, and significantly lower attrition. The mechanism, established through Self-Determination Theory and a range of randomised trials, is that autonomous motivation produces higher-quality work than controlled motivation. When the work design shifts from 'be here' to 'deliver this,' people who were previously operating below capacity because the rigid design didn't fit their lives start operating at full capacity. That's a productivity gain, and a wellbeing one.

The power dimension is the least developed of the three, and in some ways the most important. Ely and Thomas are explicit about it: the learning-and-effectiveness paradigm only operates when power is genuinely shared enough that people from different backgrounds can actually influence the work. A diverse team in which the senior leader's view always prevails - or in which people from underrepresented groups have learned not to surface the perspective that makes them different - isn't a diverse team in any operationally meaningful sense. It's a demographically varied team running on the assumptions of whoever is most powerful in the room.

The Gartner research on accountability quantifies this from a different angle. Organisations where business leaders - not HR, not the DEI team - hold high accountability for inclusion show 49% higher inclusion scores, 40% higher engagement and 14% higher employee performance than those where accountability is low. That 14% is a power finding that measures what happens when people who control decisions - budgets, promotions, performance ratings - are held responsible for whether the people around them can contribute fully. The difference between an organisation where that accountability exists and one where it doesn't is almost entirely about how power is held and used.

A Different Model

The model that the research supports isn't 'do more DEI.' It's 'do different DEI, aimed at different things.' Stop anchoring the investment in awareness training that the evidence says changes attitudes weakly and behaviour barely. Stop treating representation targets as the outcome rather than a downstream indicator. Start investing in the three mechanisms that actually connect to performance: the inclusion conditions that allow diverse perspectives to be heard and used; the flexibility that enables people who were previously constrained to contribute at full capacity; and the accountability structures that put genuine power over outcomes in the hands of the people who make talent decisions.

There is one warning worth naming. Georgeac and Rattan's research, published in the Journal of Personality and Social Psychology in 2023, found that framing DEI primarily as a business case - 'diversity is good for profit' - can itself backfire with the people the programs are meant to support. When underrepresented groups see their inclusion justified primarily in instrumental terms, it lowers their anticipated sense of belonging. The fairness argument and the performance argument are both valid. The performance case is the one that persuades sceptical boards and CFOs, but it needs to sit alongside a genuine commitment to fairness, not replace it.

In practical terms, the shift looks like this. Boards asking 'how much are we spending on DEI?' and 'what are our representation numbers?' are asking the activity and output questions. The performance questions are different: 'where in our organisation is psychological safety lowest, and what is the decision quality like in those teams?' 'Which parts of our workforce are operating below capacity because the work design doesn't fit them?' 'Which of our line leaders actually hold - and are held accountable for - the conditions that determine whether the people around them can do their best work?' Those are harder questions. They're also the ones with measurable answers, and they connect directly to the performance outcomes the conventional DEI model has spent decades failing to move.

The evidence that diversity improves performance when the enabling conditions are present has been available for years. The reason most organisations haven't acted on it is that the enabling conditions require changing how power operates, not just who sits in the room.

The Bottom Line

The reason DEI hasn't delivered the performance gains it promised is not that inclusion doesn't drive performance. It's that training and targets don't create inclusion. What creates inclusion is psychological safety - which is a function of how power is exercised in the room. What sustains it is flexibility - which is a function of who controls the conditions of work. The organisations that have understood this are running the most effective DEI programs.  They’re aimed at the conditions that determine whether diverse talent can actually do its best work - and that is a measurably different thing from what most DEI budgets are currently buying.

SOURCES & FURTHER READING

• Cassells R, Duncan A et al. BCEC/WGEA Gender Equity Insights 2020: Delivering the Business Outcomes

• Chang EH, Milkman KL, Gromet DM, Rebele RW, Massey C, Duckworth AL, Grant A. 'The Mixed Effects of Online Diversity Training.' PNAS 116(16), 2019

• Edmondson A. 'Psychological Safety and Learning Behavior in Work Teams.' Administrative Science Quarterly 44(2), 1999

• Ely RJ, Thomas DA. 'Getting Serious About Diversity: Enough Already with the Business Case.' HBR, Nov-Dec 2020

• Gartner, 'HR Leaders Must Establish Consequential Accountability to Achieve Diverse Leadership Benches' (June 2021)

• Georgeac OAM, Rattan A. 'The Business Case for Diversity Backfires.' Journal of Personality and Social Psychology 124(1), 2023

• Google re:Work, 'Understand Team Effectiveness'

• Green J, Hand J. 'McKinsey's Diversity Matters/Delivers/Wins Results Revisited.' Econ Journal Watch 21(1), 2024

• Larcker D et al. 'The Limited Corporate Response to DEI Controversies.' Harvard Law School Forum on Corporate Governance (2025)