The tug-of-war over office attendance has turned into a test of will. Some leaders equate visibility with control and push people back to their desks. Workers, particularly those with options, vote with their feet. The numbers now tilt against rigid return-to-office edicts and toward a different question: what kind of power actually improves performance when work is distributed across time, tools, and teams?
What mandates really buy you
Recent surveys cut through the noise. In February 2025, only 12 percent of U.S. executives with hybrid or remote staff planned any new mandate in the next year, according to a Stanford and Atlanta Fed survey of more than 1,000 executives. If implemented, those plans would trim paid work-from-home days in the U.S. from 21.2 to 20.8 percent, a shift of less than half a percentage point (Bloom, Davis and Barrero, 2025). Workers are not shy about their preferences either. In December 2024, only 44 percent of U.S. workers said they would comply with a fully in-person policy. The rest said they would resign or look for another job (Bloom, Davis and Barrero, 2025).
What happens when firms press ahead anyway? A working paper by Ding and Ma uses distributional synthetic controls on 260 million résumés to estimate the causal impact of mandates at Microsoft, SpaceX, and Apple. The pattern is consistent and costly. After Microsoft’s April 2022 policy, average tenure at the top end fell by roughly two months, with a 6 percent increase in the mass of employees with zero to three years at the firm. Apple’s lighter policy of one day a week shifted the seniority mix by under 4 percent. Microsoft’s 50 percent presence requirement moved it by more than 4 percent. SpaceX’s full-time requirement produced the largest effect. The stronger the mandate, the bigger the senior drain (Ding and Ma, 2024). The study is not yet peer-reviewed, but its design and dose-response result make it the clearest causal signal available.
The practical read is simple. Mandates are a blunt tool that risks your institutional memory. Veterans are often the ones who know the exception paths, the customer history, and the workarounds that never made it into the deck. Pushing them out to win a presence battle weakens the very power base leaders claim to defend.
Presence as a weak substitute for authority
Why do mandates persist if the talent math looks this bad? Because presence is an easy proxy for managerial control. It is also a poor one. Lee and Edmondson’s synthesis of self-managing organizations defines their core move as the radical, formal, and systematic decentralization of authority that eliminates the manager–subordinate relationship (Lee and Edmondson, 2017). Their review of the broader management literature underscores a boundary condition that matters here. Hierarchy performs well when executing known tasks in stable environments. It struggles with complex, cross-functional problems and tends to rigidity and status differences that stunt people’s development (Lee and Edmondson, 2017).
That is the trap with presence. In stable, repetitive work, proximity can smooth handoffs and speed supervision. In complex, interdependent work, using the office to reassert status often crowds out initiative. If the power to decide sits a level up, people learn to wait. If the sign of commitment is being seen, people optimize for being seen. Neither improves how decisions get made.
Self-management is not an ideology about place. It is a design choice about power. When decision rights are explicit, peer review is routine, and information is broadly accessible, the need to observe people minute by minute recedes. Authority lives in the process, not in the person at the head of the table. Presence can then be used for what it does best: structured collaboration and fast-cycle problem solving, not surveillance.
Safety beats surveillance
The social mechanism behind this is well established. In a field study of 51 manufacturing teams, Amy Edmondson introduced the construct of team psychological safety as the belief that it is safe to take interpersonal risks. Teams with higher psychological safety engaged in more learning behaviors, and learning mediated the link to performance (Edmondson, 1999). Put plainly, people who are not afraid to speak up solve problems faster. No attendance policy can substitute for that.
The broader engagement picture adds urgency. Global employee engagement fell to 20 percent in 2025, the lowest since 2020, with an estimated economic cost of about 10 trillion dollars, or 9 percent of global GDP (Gallup, 2025). Managers are struggling most. Their engagement has dropped nine percentage points since 2022, including a fall from 27 to 22 percent between 2024 and 2025 (Gallup, 2025). In organizations that get this right, 79 percent of managers are engaged, almost four times the global average, which suggests the gap is about organizational design, not inevitability (Gallup, 2025).
None of this proves that RTO mandates cause disengagement. The evidence does suggest that asking already strained managers to police badges will not fix output. It shifts attention from outcomes to optics and risks pushing out the very people who hold the web of tacit knowledge together. If leaders want power that lasts, they should invest in conditions that raise learning and engagement, then use the office intentionally to accelerate them.
The practical playbook
The path forward is not mysterious. It does require replacing presence as a proxy with clearer structures of authority and learning.
- State decision rights and escalation paths in writing. Who decides what, on what criteria, with what input. This is the spine of distributed authority described in the self-management literature (Lee and Edmondson, 2017).
- Treat the office as a coordination venue, not a default. Bring teams together for design sprints, retrospectives, and cross-boundary problem solving, then let them execute where they do their best work. The data suggests few executives plan to tighten further anyway, and workers will walk rather than accept blanket mandates (Bloom, Davis and Barrero, 2025).
- Protect senior talent explicitly. Before announcing any attendance rule, model its effect on tenure distribution. The strongest available causal evidence links stricter mandates to a disproportionate loss of veterans, with the effect rising with rigidity (Ding and Ma, 2024).
- Build psychological safety on purpose. Make it normal to surface risks, document experiments, and learn in public. That is the channel through which teams improve performance, whether colocated or distributed (Edmondson, 1999).
- Relieve managers of attendance policing. Aim their attention at clarity of goals and feedback. The engagement gap shows that better practice is achievable and material. The cost of getting it wrong is visible in the global productivity bill (Gallup, 2025).
The power that matters now is the ability to make and keep better decisions, faster. Counting bodies in chairs is a weak stand-in for that. Design authority well, use the office for its comparative advantage, and you keep your veterans while raising the floor for everyone else.
Sources
- Bloom, Davis and Barrero (2025). Stanford Report — Return to office? Not everybody is doing it. Survey of 1,000+ U.S. executives shows few plan new mandates and workers would often not comply, framing the limits of presence-based policy. https://news.stanford.edu/stories/2025/03/return-to-office-not-everybody-is-doing-it
- Ding and Ma (2024). Return-to-Office Mandates and Brain Drain. Working paper using distributional synthetic controls finds stricter RTO mandates at Microsoft, SpaceX, and Apple disproportionately reduced senior tenure and shifted the workforce toward less-experienced employees. https://arxiv.org/pdf/2405.04352
- Lee and Edmondson (2017). Self-managing organizations: Exploring the limits of less-hierarchical organizing. Peer-reviewed synthesis arguing that hierarchy fits stable, known tasks but inhibits cross-boundary problem solving, and defining SMOs as formal decentralization of authority. https://www.sciencedirect.com/science/article/abs/pii/S0191308517300059
- Edmondson (1999). Psychological Safety and Learning Behavior in Work Teams. Seminal field study showing team psychological safety predicts learning behaviors that mediate performance, highlighting conditions presence cannot replace. https://journals.sagepub.com/doi/10.2307/2666999
- Gallup (2025). State of the Global Workplace 2025. Global engagement fell to 20 percent with an estimated 10 trillion dollar productivity cost, and manager engagement declined sharply, underscoring the stakes of organizational design. https://www.gallup.com/workplace/659610/state-global-workplace.aspx