The RTO data disconnect: beliefs vs behavioural evidence

The return-to-office debate has become a contest between conviction and evidence, and conviction is winning.

In KPMG's 2024 CEO Outlook, 83% of the 1,345 chief executives surveyed expected a full return to the office within three years, up from 64% a year earlier.¹ The trajectory of belief is steep. The trajectory of supporting evidence is flat. A difference-in-differences study of S&P 500 firms that imposed mandates found no improvement in financial performance or firm value, alongside measurable declines in job satisfaction and in employees' ratings of senior management.² The strongest single study points the same way: the only large randomised controlled trial of hybrid work, run on 1,612 Trip.com employees and published in Nature, found that two days a week at home had no effect on performance reviews or promotion rates, while cutting quit rates by a third. Managers who had forecast a 2.6% productivity loss revised their estimate to a 1% gain once the experiment was over.³

What mandates reliably produce is not performance but departures, and of the wrong people.

The cost ledger nobody mandates for

A study of 54 S&P 500 firms that announced mandates in 2022 and 2023, tracing the employment histories of more than three million workers, found a 13 to 14% rise in abnormal turnover after the announcements, an effect that survived parallel-trends and placebo checks.⁴⁵ The departures clustered among senior, high-performing and female employees, the people hardest to replace.⁴ A separate analysis of Microsoft's mandate found a comparable pattern: senior staff left for large competitors, carrying institutional knowledge with them.⁶ Firms can improve the visible output of those who stay while quietly degrading the capability of the whole.

If the evidence is this consistent, why does the belief hold? The honest answer is that the belief is not really about productivity. The most reliable predictors of strict in-office mandates, according to a 2026 conference paper, are not economic fundamentals but firm and chief-executive characteristics: larger firms, and older and male chief executives.⁷ Leaders are also enforcing schedules they do not personally endorse. Only 16% of business leaders think a five-day office week is ideal, yet 30% of companies plan to require one by 2026.⁸ Enforcement has outrun conviction.

Underneath it sits a measurement failure. Microsoft's Work Trend Index found that 87% of employees consider themselves productive, while only 12% of leaders are fully confident their teams are.⁹ The gap is a visibility problem. Managers who cannot see outcomes fall back on the proxy they can see, which is presence, and tracking activity rather than impact then drives the very behaviour it fears, as people perform busyness to be seen working.⁹

What the office is genuinely for

None of this means the office is obsolete, and the strongest counter-evidence deserves full weight. A study of software engineers using code-review records as a paper trail found that those seated near their teammates gave and received substantially more feedback. When offices emptied, junior engineers lost the most mentorship, and women, more likely to seek feedback in person, were hit hardest.¹⁰ Proximity builds human capital even where it costs short-run output. Workers agree in their own way: they rate the office better for collaboration and the home better for focus.¹¹

The conclusion that follows is not that remote wins. It is that location is contingent. Co-location pays for specific people doing specific things at specific times: onboarding, or the feedback-heavy phases of a project. A blanket five-day mandate cannot tell the difference between an engineer who needs a mentor across the desk and a senior individual contributor whose deep work the commute simply interrupts. The proximity evidence is an argument for designed, purposeful in-person time, not for uniform attendance.

That is the kind of pattern behavioural data can detect. The productivity question, asked properly, concerns which work, for whom, produces what, and where. Collaboration cadence and output rhythm can be measured the same way wherever the work happens, which makes location a variable to test rather than a doctrine to impose. The two-year study of 800,000 employees that found leadership quality a larger productivity factor than location was making the same point: the lever leaders reach for is rarely the one that moves the outcome.¹²

Executive scepticism of self-reported productivity is not irrational, and the 87% who call themselves productive are, after all, grading their own homework.⁹ But mandate-by-intuition is the weaker position, because the outcome-based studies that exist find no mandate dividend at all. Both camps are arguing from thin signal. The way forward is less about picking a side in the office wars than about collecting the evidence that would settle the argument inside a given company. Trip.com ran the experiment before it wrote the policy. How many firms enforcing five days have run theirs?

Footnotes

1. Going Concern News Desk. (2024). CEOs surveyed by KPMG feel a full return to office is imminent. Going Concern. https://www.goingconcern.com/ceos-surveyed-by-kpmg-feel-a-full-return-to-office-is-imminent ↩

2. Ding, Y., & Ma, M. (S.). (2024). Return to office mandates don't improve employee or company performance [research summary]. University of Pittsburgh, Katz Graduate School of Business. https://www.business.pitt.edu/return-to-office-mandates-dont-improve-employee-or-company-performance ↩

3. Bloom, N., Han, R., & Liang, J. (2024). Hybrid working from home improves retention without damaging performance. Nature, 630(8018), 920–925. https://www.nature.com/articles/s41586-024-07500-2 ↩

4. Ding, Y., Jin, Z., Ma, M. (S.), Xing, B. (B.), & Yang, Y. (J.). (2025, October 8). Return-to-office mandates and the hidden cost of brain drain. Hankamer School of Business, Baylor University. https://hankamer.baylor.edu/news/story/2025/return-office-mandates-and-hidden-cost-brain-drain ↩ ↩²

5. Ding, Y., Jin, Z., Ma, S., Xing, B., & Yang, Y. (2024). Return-to-office mandates and brain drain [Working paper]. https://cdn.arstechnica.net/wp-content/uploads/2024/12/5031481.pdf ↩

6. Wright, A. (2024). Return to office and the tenure distribution [Working paper]. University of Chicago, Harris School of Public Policy. https://harris.uchicago.edu/sites/default/files/wright-return-to-office.pdf ↩

7. Determinants and consequences of return to office policies. (2026). [Preliminary conference paper]. American Economic Association Annual Meeting. https://www.aeaweb.org/conference/2026/preliminary/paper/hFn8f62s ↩

8. Fair Play Talks. (2025, October 16). Three in ten companies plan to end remote work & enforce return to office by 2026, survey finds. Fair Play Talks. https://www.fairplaytalks.com/2025/10/16/three-in-ten-companies-plan-to-end-remote-work-enforce-return-to-office-by-2026-survey-finds/ ↩

9. Microsoft WorkLab. (2022). Hybrid work is just work. Are we doing it wrong? (Work Trend Index Special Report). Microsoft. https://www.microsoft.com/en-us/worklab/work-trend-index/hybrid-work-is-just-work ↩ ↩² ↩³

10. Emanuel, N., Harrington, E., & Pallais, A. (2023). The power of proximity to coworkers: Training for tomorrow or productivity today? (NBER Working Paper No. 31880). National Bureau of Economic Research. https://www.nber.org/system/files/working_papers/w31880/w31880.pdf ↩

11. Owl Labs. (2025). State of hybrid work 2025: US report. Owl Labs. https://owllabs.com/state-of-hybrid-work/2025 ↩

12. Founder Reports. (2026). Essential return-to-office statistics and trends (2026). Founder Reports. https://founderreports.com/return-to-office-statistics/ ↩