Mandates have returned workers to their desks. Engagement has not followed. In 2025 global employee engagement fell to 20 percent, the lowest since 2020, with an estimated 10 trillion dollars in lost productivity, roughly 9 percent of global GDP (Gallup, 2025). The picture is worse for managers, whose engagement slid to 22 percent after the largest annual drop between 2024 and 2025. If the office were the cure, these numbers would be moving the other way.
The venue is not the variable. Voice is. The question is whether people can surface problems, test ideas, and admit uncertainty without fear. That is psychological safety, a team-level belief that it is safe to take interpersonal risks. The link to performance is not a slogan. It rests on a quarter century of research across industries and countries.
What actually changes team performance
Amy C. Edmondson first formalized team psychological safety and tested it in a field study of 51 work teams inside a single manufacturing company (Edmondson, 1999). Teams with higher psychological safety showed more learning behaviors, such as asking for help, discussing errors, and seeking feedback. Those behaviors, in turn, explained the teams’ performance. Team efficacy did not predict learning once safety was accounted for. The design was cross-sectional and within one firm, so it is not a randomized trial, but the mechanism it identified has held up: safety enables learning, and learning drives outcomes.
A decade and a half later, a comprehensive meta-analysis aggregated 136 independent samples covering more than 22,000 individuals and nearly 5,000 groups (Frazier et al., 2017). The corrected correlations are hard to ignore. Psychological safety is strongly related to learning behavior (ρ̂ = .62), information sharing (ρ̂ = .52), engagement (ρ̂ = .45), and task performance (ρ̂ = .43). The link to creativity is positive but smaller (ρ̂ = .13). These are not causal estimates. They do show, across many settings, that where people believe it is safe to speak up, teams report more of the very behaviors leaders say they want when markets turn volatile.
The design features leaders can move
The same meta-analysis identified job design antecedents that track with psychological safety (Frazier et al., 2017). Autonomy correlates at ρ̂ = .47, role clarity at ρ̂ = .63, and a supportive context at ρ̂ = .49. Interdependence shows a very large individual-level correlation of ρ̂ = .86, but that estimate is based on only five samples and is more modest at the group level (ρ̂ = .40). The lesson is practical. You do not buy safety with platitudes. You design for it.
- Clarify roles so people know where their decision rights begin and end. Ambiguity breeds caution.
- Grant real autonomy over work methods and sequencing. Discretion without retribution signals trust.
- Build visible support into the context, from access to information to coaching time.
- Be thoughtful about interdependence. Tightly coupled work can heighten the need for coordination and voice, but do not overinterpret one large coefficient built on a small base. Treat it as a hypothesis to test in your setting.
These levers translate directly to hybrid and distributed models. If you insist on physical presence but leave decision rights fuzzy, you will get compliance theater, not collaboration. If you give teams clear guardrails and latitude to solve problems, you get the behaviors the evidence links to performance.
RTO, manager fatigue, and the hierarchy trap
Why is this design work urgent now? Because the people who set local conditions, managers, are themselves disengaged. Manager engagement fell to 22 percent in 2025, and the year-over-year drop from 27 to 22 percent was the steepest in the recent series (Gallup, 2025). This is descriptive survey evidence, not a causal claim, but it points to a system under strain. In organizations that operate as exemplars, 79 percent of managers are engaged, almost four times the global average, which suggests design choices separate the best from the rest (Gallup, 2025).
The structure of authority matters as much as the setting of work. Hierarchical management excels at reliable execution of known tasks in stable environments. It falters when problems are complex, cross functional, and evolving, and it tends to magnify status differences that suppress development and voice (Lee and Edmondson, 2017). Self-managing organizations push decision authority down systematically and eliminate the classic manager–subordinate relationship. That choice raises the premium on norms and processes that make it safe to raise a hand. You cannot decentralize judgment if people fear looking foolish in front of peers.
Return-to-office policies aim to restore coordination and culture by proximity. That can help with spontaneity. It does not neutralize the status dynamics or fear that keep errors hidden. If anything, putting people back in rooms without redesigning roles and rituals can intensify impression management. The research record points to a different operating move: reduce unnecessary hierarchy in the work, make decision rights explicit, and build routines for candor, then use the office as a tool, not a talisman.
The operating move
Leaders should treat psychological safety as a measurable team condition and manage it like any other performance input. Start with job design changes the meta-analysis flags: sharpen role clarity, expand autonomy where risk is tolerable, and strengthen visible support systems (Frazier et al., 2017). Pair that with team routines that institutionalize learning behaviors identified in the original field study, such as structured debriefs after projects and explicit invitations for dissent before decisions lock (Edmondson, 1999). In more decentralized settings, make the structural choice explicit. If you remove layers, you must add forums where peers can question each other’s decisions without penalty, or the benefits of decentralization will evaporate under quiet conformity (Lee and Edmondson, 2017).
None of this requires waiting for a new survey cycle. A senior team can rewrite one meeting agenda this week to surface risks without blame. A product leader can publish decision rights for the next quarter. An operations manager can carve out an hour for peer coaching tied to the thorniest cross-boundary problem. These moves do not guarantee engagement will surge. They do make it more likely, because they target the social conditions that the evidence associates with learning, information flow, and performance.
The practical implication is stark. If your RTO plan does not include clear steps to raise psychological safety, it is a real estate strategy passing as culture. The cost of that mistake is already visible in the engagement numbers. The upside is visible too. Where managers operate in well-designed systems, engagement climbs toward best-practice levels, and the trust to speak up becomes the engine of performance, not a by-product of presence (Gallup, 2025; Frazier et al., 2017; Edmondson, 1999; Lee and Edmondson, 2017).
Sources
- Edmondson (1999). Psychological Safety and Learning Behavior in Work Teams. Introduced team psychological safety and showed in a single-firm field study that safety predicts learning behaviors, which in turn explain team performance. https://journals.sagepub.com/doi/10.2307/2666999
- Frazier, Fainshmidt, Klinger, Pezeshkan and Vracheva (2017). Psychological Safety: A Meta-Analytic Review and Extension. Aggregated 136 samples and found strong correlations between psychological safety and learning, information sharing, engagement, and task performance, plus job design antecedents with noted caveats. https://doi.org/10.1111/peps.12183
- Lee and Edmondson (2017). Self-managing organizations: Exploring the limits of less-hierarchical organizing. Synthesizes evidence that hierarchy suits routine execution but impedes complex cross-boundary problem solving and elevates status barriers to voice, clarifying what decentralization demands. https://www.sciencedirect.com/science/article/abs/pii/S0191308517300059
- Gallup (2025). State of the Global Workplace 2025. Reports engagement at 20 percent, about 10 trillion dollars in lost productivity, steep manager engagement declines to 22 percent, and best-practice manager engagement at 79 percent, indicating design gaps. https://www.gallup.com/workplace/659610/state-global-workplace.aspx