Every return-to-office mandate issued since 2023 has come wrapped in the same language: rebuilding culture, sharpening collaboration, restoring the serendipity of hallway conversations. Almost none of that language survives contact with the data. What the last three years of RTO enforcement actually reveal has nothing to do with culture and everything to do with something CEOs are far less comfortable measuring: who, precisely, holds the leverage in their workforce, and what happens when that leverage is tested at scale.
The cleanest evidence comes from a study out of Baylor University’s Hankamer School of Business, which tracked 54 large S&P 500 technology and financial firms that issued RTO mandates between 2020 and 2023, drawing on LinkedIn-based employment records for more than three million workers. The finding was not that people quit in large numbers. Aggregate attrition barely moved. The finding was about who quit. Abnormal turnover rose 13 to 14 percent following mandate announcements, but the increase was concentrated almost entirely among the employees a company can least afford to lose: mid- and senior-level managers, high-skilled specialists, and women at nearly three times the rate of men. Job vacancies left by departing staff took 23 percent longer to fill. Hiring rates fell 17 percent. A separate University of Pittsburgh analysis of the same cohort found job satisfaction fell sharply while firm-level financial performance showed no corresponding gain, which is the quiet part that rarely makes it into the board memo: the policy didn’t even deliver the productivity case its authors used to justify it.
The Adverse Selection Problem
What’s happening here has a name in economics, borrowed from insurance markets: adverse selection. When you impose a uniform cost on a heterogeneous population, the people who leave first are not a random sample. They’re the ones with the best outside options. In a tight, still-competitive market for senior engineers, product leaders and skilled analysts, a blanket in-office requirement doesn’t filter for commitment or performance. It filters for mobility. And mobility correlates far more closely with talent than most RTO memos assume.
Amazon’s own internal numbers make the mechanism visible in real time. After the company’s five-day mandate took effect in 2025, an internal survey recorded a satisfaction score of 1.4 out of 5, with more than nine in ten respondents expressing dissatisfaction. By that November, 48 percent of affected employees said they had already applied elsewhere, and more than two-thirds expected to leave within the year. Competitors treated the mandate as a recruiting opportunity. Meanwhile the rollout itself exposed a company that hadn’t planned for its own policy: seven cities lacked enough desks, some needing 30 percent more capacity than existed, and Seattle commute times rose by up to a fifth. Dell ran a parallel experiment with a sharper edge, tying promotion eligibility directly to office attendance. Nearly half of its eligible U.S. workforce chose to forgo advancement rather than comply, a result that only makes sense if you assume a meaningful share of Dell’s talent believed it had better alternatives than the promotion on offer.
None of this means remote work is inherently more productive, a claim the evidence only partially supports. It means the RTO decision is not being priced correctly. Executives keep asking whether the policy improves culture. The data keeps answering a different question: what fraction of your best people can walk, and will they?
Where the Same Policy Produces the Opposite Result
This is where the picture gets genuinely interesting for anyone running a global organisation, because the same mandate, issued in a different labour market, produces close to the opposite outcome. India’s largest IT services firms have run RTO policy at least as aggressively as their American clients. TCS moved to a five-day requirement and later tightened work-from-home allowances to six days a quarter. Infosys instituted a minimum of ten office days a month for the bulk of its workforce, covering roughly 323,000 employees. Wipro tied attendance directly to compensation. Yet none of these firms have reported anything resembling Amazon’s flight of talent or Dell’s promotion boycott. What they have produced instead is a commercial property boom: Bengaluru absorbed 6.4 million square feet of office space in a single quarter of 2026, up 13 percent year-on-year, prime rents have risen roughly 30 percent, and the country’s national office supply-demand gap has widened to around 9 million square feet, with global capability centres now accounting for more than 40 percent of total demand.
The difference isn’t cultural. It’s structural. A large share of India’s IT workforce operates in a labour market with far less individual bargaining power than their counterparts in Seattle or Austin, a condition sharpened for the segment of that workforce deployed onshore in the United States, where tightening scrutiny of the H-1B programme, including a steep new petition fee introduced by the Trump administration, has made changing employers materially riskier for visa-dependent staff. TCS’s own five-day mandate for U.S. employees, issued in mid-2025, landed on a workforce with unusually little room to push back. The policy that triggers an exodus in one market becomes, in another, simply the price of employment. Companies running both playbooks side by side are not testing one policy. They’re running a live comparison of bargaining power across geographies, and mistaking the results for a verdict on office culture.
What This Should Change in the Boardroom
Three implications follow, and none of them are the usual hybrid-versus-office debate. First, any board using RTO as an unstated cost-reduction lever, hoping natural attrition will do the work layoffs would do more visibly, should understand it is running adverse selection against its own bench. The people most likely to self-select out are precisely the ones with the resume and the network to land elsewhere quickly, which is rarely the underperforming quartile leadership actually wants to thin. Second, the metric that matters is not the attrition rate, it’s the quality-weighted attrition rate. A company that loses 4 percent of headcount but 12 percent of its top-decile technical talent has a governance problem that a flat retention dashboard will not surface, and few HR reporting systems currently disaggregate departures by performance tier for the people approving these policies. Third, the geographic arbitrage that made global capability centres and offshore hubs attractive in the first place, cheaper, flexible, high-quality labour, is being partly eroded by the real estate inflation that aggressive in-office mandates are now generating in those same hub cities. A strategy built on cost advantage should not be indifferent to the fact that Bangalore office rents just rose 30 percent in a year substantially because of RTO-driven demand.
There is one legitimate case for in-person requirements that has nothing to do with nostalgia for hallway conversations: as agentic AI systems take on more first-line decision-making, organisations may need tighter physical coordination for the judgment calls and error escalation that still require humans in a room together, quickly. That is a specific, testable claim about workflow, and it can be evidenced with cycle-time and error-correction data. It is a different claim from “we want you back for culture,” and CEOs who conflate the two will keep making policy on the wrong evidence. The honest question a board should be asking before its next mandate is not whether the office is good for culture. It’s a harder one: which of my people have somewhere else to go, and am I prepared to find out the hard way which of them actually leave.
