The debate over office attendance is unusually confident for a question with so little clean evidence behind it. Advocates for the mandate cite collaboration and culture; opponents cite autonomy and focus. What almost no one cites is a randomized study of a return-to-office mandate, because there barely is one. The best causal research measures something adjacent: what happens when you let people work from home. Read carefully, it neither vindicates the mandate nor promises that remote work is free.
The clean experiments
The cleanest evidence comes from two randomized trials at Chinese firms. In the first, run at the travel agency Ctrip and published in 2015, call-center staff who volunteered were randomly assigned to work from home or from the office for nine months. Home workers were about 13% more productive, roughly nine points of that from working more minutes per shift and about four from a quieter environment. But there was a catch that both camps tend to skip: their promotion rate, conditional on performance, fell. Being out of sight had a career cost even as it lifted output.
The second trial, published in Nature in 2024, tested hybrid rather than full remote. More than 1,600 graduate employees at Trip.com were randomized to two days a week at home or to full office attendance. Hybrid work cut voluntary resignations by about a third, from 7.2% to 4.8%, and equivalence tests found no difference in performance-review scores across four cycles and no difference in promotions over the following two years. Among the software engineers in the sample, lines of code did not move. Two days at home, in this experiment, bought lower turnover at no measurable cost to output.
What the studies do not say
Here the honesty has to run in both directions. Neither trial studied a mandate to return; they studied the freedom to leave. The direction of the intervention matters, because a policy that grants autonomy and a policy that revokes it are not mirror images. Nor are these results universal. Both firms are large, Chinese, and staffed by educated professionals; a call center and a travel company are not a design studio or a hospital. The 2015 promotion penalty is a real finding, and it warns that fully remote arrangements can quietly stall careers even when they raise measured output.
The pattern that survives scrutiny is not "remote wins" or "office wins." It is that the arrangement's design carries the result. When Ctrip later opened the choice to everyone and let the original participants reselect, the productivity gain nearly doubled, to about 22%, because people sorted into the arrangement that fit them. Choice, not location, did the work. That is a finding about matching people to conditions, which is a management problem, not a real-estate one.
Design over presence
When the same researchers translated their hybrid results for managers, they did not conclude "send everyone home." They named three conditions for hybrid to succeed: rigorous performance management, coordinated schedules so teams overlap, and genuine support from senior leadership. Every one of those is about how work is organized, not where it sits. A hybrid policy without them can fail; a mandate with none of them can look like it works only because no one is measuring output.
That is the uncomfortable center of the evidence. The studies that get quoted in support of the office are largely silent on mandates, and the studies that measure productivity keep pointing back to autonomy, matching, and coordination. A leader who wants to use the research honestly has to give up the clean slogan. The defensible claim is narrow: some tasks benefit from shared time and space, hybrid arrangements can protect both output and retention when they are designed well, and full remote work trades focus and flexibility against a real risk to visibility and advancement. Anyone claiming more than that is borrowing the credibility of studies they have not read to the end.
Sources
- Bloom, Liang, Roberts & Ying (2015). Does Working from Home Work? Evidence from a Chinese Experiment (QJE). RCT: WFH raised performance ~13% but cut promotion rate conditional on performance; with company-wide choice, gains rose to ~22%. https://academic.oup.com/qje/article-abstract/130/1/165/2337855
- Bloom, Han & Liang (2024). Hybrid working from home improves retention without damaging performance (Nature). RCT: two-day hybrid cut resignations ~a third (7.2%→4.8%, P=0.043) with no measured hit to performance or promotion. https://www.nature.com/articles/s41586-024-07500-2
- Bloom, Han & Liang / HBR (2024). One Company A/B Tested Hybrid Work. Three conditions for success: rigorous performance management, coordinated schedules, leadership support. https://hbr.org/2024/10/one-company-a-b-tested-hybrid-work-heres-what-they-found