For a war, remarkably little territory is changing hands. A Stanford and Atlanta Fed survey of more than 1,000 executives in February 2025 found that only 12% of those with hybrid or remote workers planned any return-to-office mandate in the following twelve months, and that the planned returns would move the share of paid days worked from home in the US from 21.2% to 20.8%: less than half a percentage point. Remote work, per the same research team's NBER estimates, has settled at roughly a quarter of all paid workdays in the US. The headlines describe a great reversal. The data describe an equilibrium, defended by employers themselves, most of whom told the surveyors they would not change course even in a recession.
So why does the fight feel existential? Because it is not actually about where people sit. It is about three older questions that the office debate lets everyone argue by proxy.
The productivity question is mostly answered
Start with the stated battlefield, because it is the least contested. The strongest single piece of evidence is a randomized trial at Trip.com covering 1,612 professionals: hybrid work (two days at home) cut attrition by a third, 7.20% quit rate in the control group versus 4.80% in the treatment group, with no detectable difference in performance ratings across four semiannual review cycles over two years, and no difference in promotions. Among the 653 software engineers in the sample, even lines of code showed no effect. This is what settled science looks like in management: a real RCT, in a real firm, with objective outcomes.
Executives largely know this, which is why so few plan mandates and why the ones who impose them rarely cite productivity data. When a fight continues after the empirical question closes, the fight was never empirical.
What the war is actually over
The first real issue is control, and here the evidence turns diagnostic. Ding and Ma, applying synthetic-control methods to 260 million résumés, found that RTO mandates at Microsoft, SpaceX and Apple triggered disproportionate departures of senior employees, with a dose-response pattern: the stricter the mandate, the larger the seniority drain, from under 4% of the distribution at Apple (one day a week) to as much as 15% at SpaceX (full-time presence). The seniors did not leave for startups; they went to large direct competitors. A policy that repels precisely the people with the most options is not a productivity policy. It is a loyalty test, and loyalty tests select for people who cannot afford to fail them.
The second issue is what management is for. In the office, a manager can supervise inputs: presence, hours, visible effort. Distributed work forces supervision of outputs, which is harder, lonelier, and exposes which managers were adding judgment and which were adding attendance. Some portion of mandate energy is, frankly, a middle layer defending its own legibility, at the exact moment AI is already squeezing that layer from the other side.
The third issue is trust, and it is the oldest. Ricardo Semler saw this pattern four decades before the pandemic, as he has told it, in the strange spectacle of Semco having to make vacations compulsory: thirty days, out, laptop locked away, because adults kept negotiating to work more. His diagnosis was exact: that is not loyalty, it is fear, "the cheapest management tool ever invented." Fear-based presence is cheap because it requires no measurement, no clarity about outcomes, no actual management. The mandate is its modern form: a policy that replaces the difficult question "is the work good?" with the easy one "is the body here?"
The tell in December's data
One number gives the game away. In the December 2024 Survey of Working Arrangements and Attitudes, only 44% of American workers said they would comply with a full-time RTO policy; the rest said they would quit or start looking. Employers can read surveys too, which is why 88% plan no mandate: they understand the current arrangement as a wage. Flexibility is compensation, and clawing back compensation from your most mobile employees is a strategy with a known ending, now documented at three of the most prestigious employers in America.
Proxy wars end when the real dispute gets named. Named plainly: this is a negotiation over whether management runs on measurement or on monitoring, and over who bears the cost of managerial convenience. A leadership team that wants people together for genuinely collaborative work can usually get them, by making the office worth the commute and saying precisely what the days together are for. A leadership team that wants presence as proof of submission can get that too, and will, per the evidence, receive exactly what fear buys: the people who stayed because they could not leave, managed by the metric that was easiest to check.
Sources
- Bloom, Han & Liang (2024). Hybrid working from home improves retention without damaging performance. Nature. RCT, 1,612 employees at Trip.com: quits cut by a third (7.20% vs 4.80%), no performance or promotion differences over two years. https://www.nature.com/articles/s41586-024-07500-2
- Stanford Report / SWAA (2025). Return to office? Not everybody is doing it. Only 12% of executives planned any RTO mandate; planned returns move US WFH share from 21.2% to 20.8%; Dec-2024 SWAA: only 44% of workers would comply with full-time RTO. https://news.stanford.edu/stories/2025/03/return-to-office-not-everybody-is-doing-it
- Ding & Ma (2024). Return-to-Office Mandates and Brain Drain. arXiv 2405.04352. Synthetic controls over 260M résumés: RTO at Microsoft/SpaceX/Apple caused disproportionate senior exits with dose-response by mandate strictness (up to 15% at SpaceX). https://arxiv.org/pdf/2405.04352