In 2020, when the Securities and Exchange Commission rewrote its disclosure rules to require public companies to reveal more about their human capital practices, the intent was straightforward: give investors better information. But what actually happened tells a more subtle story about how regulation reshapes corporate behavior—and how the gap between talking about change and achieving it can be surprisingly wide.
Daniele Macciocchi, an associate professor of accounting at the University of Miami Patti and Allan Herbert Business School, suspected the rule might do something else entirely. Working with Jung Ho Choi of Stanford University and Dan Li of Singapore Management University, he set out to examine whether the new disclosure requirements had nudged companies to change how they recruited talent. The team analyzed more than 5 million job postings from both public and private companies, using artificial intelligence and large language models to evaluate not just what firms said about diversity, equity and inclusion, but how genuinely they said it.
The results were striking in their nuance. After the SEC's Regulation S-K amendment took effect, public companies did indeed increase diversity-related messaging in their job postings compared with similarly sized private firms not subject to the disclosure requirement. The effect was especially pronounced among companies already facing pressure from environmental, social and governance-focused investors to improve their workforce diversity. Companies had clearly gotten the message: if you're going to discuss diversity in your annual filings, make sure you're also talking about it when you recruit.
But here's where the story gets complicated. The researchers discovered something that modern business often struggles to admit: changing what you say is far easier than changing what you do. When Macciocchi's team used AI to look beyond simple keyword matching and evaluate the actual substance of DEI messaging, they found a telling pattern. Companies that relied on generic, boilerplate language about diversity showed no actual improvement in hiring diversity. Those genuinely committed to the effort looked different in their job postings—they explained why diversity mattered to them, not just that it did.
Even among firms making authentic efforts, the structural challenges proved daunting. Companies trying to improve workforce diversity interviewed more candidates, took longer to fill positions, extended more job offers and experienced higher rejection rates from prospective employees. "Structural changes take a long time," Macciocchi noted. "Even after several years, the changes are not as large as people might expect."
The research, published in the Journal of Accounting Research, points to a broader ecosystem of change unfolding across business. Regulation, investor pressure and labor-market competition are all reshaping how firms communicate with shareholders, employees and job seekers. The study also demonstrates how AI and large language models have become essential tools for understanding corporate behavior at scale—work that would have been nearly impossible using older text-analysis methods.
For companies still figuring out their diversity commitments, the takeaway is clear: the talk has to match the walk. And that walk, as it turns out, is a longer journey than most expected when they first started talking about it.