More than a year after the governmentwide return-to-office order stripped telework agreements out of the executive branch, Federal News Network asked federal employees how it went. The answer, from 7,463 respondents surveyed in May: badly. About 76% said they are somewhat or much less productive than they were before returning. Roughly 21% said their productivity is unchanged. Three percent said they got better.
The rest of the file reads the same way. Eighty-three percent described the overall experience as negative. Fifty-seven percent were dissatisfied with the physical workspace they were ordered back into. And when asked what single change would most improve their working life, 93% named a more flexible schedule and fewer days in the building.
The caveat matters and the publication states it plainly: this is a non-scientific survey of self-selected respondents. People who hate a policy answer surveys about that policy. Self-reported productivity is also not output — it is a feeling about output, and a demoralized workforce reports demoralized numbers. Treat 76% as a temperature reading, not a measurement.
The peer-reviewed work points the same direction with more care. A 2026 study in Public Personnel Management by Geneviève Morin, Carey Doberstein and Étienne Charbonneau analyzed three waves of federal workforce survey data from 2024 and 2025 and found the January 2025 mandate disrupted established routines, cut employees' sense of control over their work environment, and added emotional and logistical strain. Not "workers are annoyed." Loss of autonomy, which is one of the better-evidenced predictors of disengagement in the literature.
Our take: The striking thing is not that the survey is negative. It is that a policy affecting hundreds of thousands of people was implemented with no before-and-after measurement, so the only evidence anyone has is a feelings poll taken eighteen months later. Compare that to the four-day-week trials, where 141 companies across six countries ran a six-month protocol on 2,896 employees and published the results. Whatever you think of the answer, one side ran an experiment and the other ran a memo.
That contrast is the real story. The largest controlled trial of a four-day week, published in Nature Human Behaviour, found worker wellbeing improved while productivity held steady or rose. Boston College's analysis of the pilot recorded a 71% drop in burnout and a 57% fall in turnover, with 90% of participating companies continuing after the trial. UK pilot revenue rose 1.4% on average during the period. There was even a dose-response curve: employees whose weekly hours fell by eight or more showed larger gains in burnout, job satisfaction and mental health than those with smaller cuts.
Those trials have their own problems — volunteer companies self-select into them, and much of the outcome data is also self-reported. But they establish a baseline, run a defined intervention, and measure the same people afterward. Almost no RTO mandate has done that.
Meanwhile the mandates keep landing. California's order for state workers took effect July 1, moving them to four in-person days a week from two, over protests about commutes, childcare and cost. Roughly 30% of companies now run fully in-person. The stated rationale is usually productivity and collaboration; the supporting evidence is usually an executive's intuition and a downtown lease.
What to watch
- Attrition, not sentiment. Surveys are noisy. Quit rates and time-to-fill in mandated agencies are not.
- Whether anyone instruments the next one. The first large employer to publish pre- and post-mandate output data will set the terms of this argument for a decade.
- The four-day dose-response. If bigger hour cuts keep producing bigger gains, the debate stops being about where you work and starts being about how much.
- State-level spread. California's four-day in-person rule is a template other states will copy or reject in public.
If you manage people, the transferable lesson has nothing to do with offices. Any change to how a team works — location, hours, tooling, meeting load — is an experiment. Run it like one. Pick two or three output measures before you change anything, hold them for a quarter, then decide. Otherwise the only data you will ever have is how everybody felt about it, collected long after the decision was already irreversible.