← Articles
Read in another language
Society

Return-to-Office Mandates: What the Evidence Actually Says

Jayden

Analyzes global supply chains, industrial policy, and technology issues.

Published

Key points

  • Amazon, JPMorgan Chase, Dell, AT&T and the US federal government all moved to full-time or near-full-time on-site work in 2025, with mandates taking effect between January and March.
  • Employer rationales — mentoring, culture, brainstorming — are stated beliefs, not measured productivity results; no mandating employer has published outcome data.
  • The strongest causal evidence, a randomised trial of 1,612 Trip.com employees published in Nature, found hybrid work cut quit rates by about a third with no significant hit to performance, promotions or code output.
  • Studies of S&P 500 mandates and of roughly 260 million resumes link RTO orders to lower job satisfaction and shorter average tenure, with senior staff leaving for direct competitors, and no measurable gain in firm value.
  • Survey data put working from home at roughly a quarter of paid workdays in 2025 with hybrid attendance clustering at three days on-site, suggesting the equilibrium — not the mandate headline — is where this settles.

On January 2, 2025, Amazon's roughly 350,000 corporate employees were expected back at their desks five days a week — the end of the hybrid schedule the company had run since 2023 [source: Amazon, 2024]. Weeks later, JPMorgan Chase ordered its own staff back full-time from March, and on his first day in office the US president directed the entire federal workforce to end remote work [source: Reuters, 2025] [source: Federal Register, 2025]. After four years in which "where you work" seemed permanently loosened, 2025 became the year a wave of employers tried to snap it shut. The question underneath the memos is narrower and more testable than the headlines suggest: when a company forces everyone back, does it actually work?

This is a story about where work happens — office versus home — and who holds the leverage to decide. It is not about working fewer hours or a four-day week, which is a debate about working time. The distinction matters, because the evidence on location has become unusually good, and it does not line up neatly with either the "return to normal" or the "remote forever" camp.

The mandate wave of 2025

For most of 2021 through 2024, the compromise was hybrid: a few days in, a few days out. In 2025 a cluster of large employers replaced that compromise with hard five-day rules. Amazon led, framing the change as a return to how things worked before the pandemic [source: Amazon, 2024]. JPMorgan followed, telling employees companywide to come in five days from March; the bank noted that more than half its people were already in the office full-time and gave the rest at least 30 days' notice [source: Reuters, 2025]. Dell told staff living within about an hour of an office to be on-site five days a week from March [source: The Boston Globe, 2025]. AT&T moved corporate staff to five days from January; Sweetgreen pushed non-frontline workers to four [source: The Boston Globe, 2024].

The public sector escalated too. On January 20, 2025, a presidential memorandum directed federal agency heads to "terminate remote work arrangements and require employees to return to work in-person… on a full-time basis," subject to exemptions [source: Federal Register, 2025]. For context, only around 10% of federal workers had been fully remote, and most of the 2.3-million-strong civilian workforce already worked in person — a reminder that "remote work" and "the office debate" were always about a specific slice of mostly white-collar jobs, not the whole economy.

The employers' stated reasons are consistent and, on their face, reasonable: collaboration, culture, mentoring, and speed. Amazon's Andy Jassy argued that being together makes it "easier for our teammates to learn, model, practice, and strengthen our culture" and that "collaborating, brainstorming, and inventing are simpler and more effective" [source: Amazon, 2024]. JPMorgan's leaders wrote that "being together greatly enhances mentoring, learning, brainstorming and getting things done" [source: Reuters, 2025]. These are genuine beliefs held by experienced managers. They are also, importantly, claims — statements of conviction about benefits, not measured results. Keeping that line clear between what leaders believe and what studies measure is the whole task of reading this debate honestly.

Two views, fairly stated

Workers who push back are not simply resisting effort. The costs they cite are concrete: commuting time and money, lost autonomy, and schedules that are harder to combine with caregiving. And the burden is not evenly distributed. Long commutes, childcare, eldercare, and disability accommodations fall unequally, which is why flexibility tends to matter more to some groups than others — a pattern the evidence below makes measurable rather than anecdotal.

But the employer side has a real point too, and it deserves to be stated at full strength rather than caricatured. Offices are not just overhead; proximity does something for the parts of work that are hard to schedule — the overheard problem, the quick correction, the apprenticeship of a junior employee watching a senior one. The honest version of this debate is not "bosses want control" versus "workers want comfort." It is a genuine tension between individual flexibility and the collective, hard-to-measure value of being in the same room. The useful question is which of those claims survives contact with data.

What the evidence actually shows

Start with the single best measurement we have. In a six-month randomized controlled trial — the gold standard, rarely achieved in workplace research — Stanford economist Nicholas Bloom and colleagues randomly assigned 1,612 employees at the travel company Trip.com to either a hybrid schedule (two work-from-home days a week) or fully in-office, then tracked them for two years [source: Nature, 2024]. Hybrid work cut quit rates by about a third and had no significant effect on performance grades, promotions, or, for engineers, lines of code written. The retention gains were largest for women (a 5.0-percentage-point drop in attrition), non-managers, and employees with long commutes [source: Nature, 2024]. In other words, for a hybrid schedule, the productivity fear was not borne out and the retention benefit was real.

Here the distinction between perceived and measured becomes vivid. Before the experiment, the company's own managers estimated hybrid work would drag productivity down by 2.6% on average; after seeing the results, they revised that to a positive 1.0% [source: Nature, 2024]. The belief that in-person time is obviously more productive is sincere and widespread — and, in this trial, it was measurably wrong for the hybrid arrangement. That is the "felt versus measured" gap at the center of the whole RTO argument.

Two cautions keep this from becoming a blanket case for working from home. First, the trial tested hybrid, not fully remote — and fully remote is genuinely more contested. A separate study of software engineers at a large firm found that sitting in the same building as teammates brought engineers about 22% more feedback on their code, and that proximity traded short-term output for long-run development, essentially mentoring; the tradeoff was sharper for women, who both gave and received more of that mentoring [source: NBER, 2023]. Proximity, in other words, has a real payoff — just one that shows up in training and human capital rather than this quarter's output. Second, one rigorous trial at one company is not the last word. But it is far stronger evidence than a memo, and it points the same direction as the broader data on retention.

The cost of forcing it

If hybrid is roughly neutral-to-positive, what happens when employers skip it and mandate five days? The most direct study looked at S&P 500 firms that imposed return-to-office rules and compared employee sentiment before and after, using a difference-in-differences design on Glassdoor reviews. It found a sharp decline in job satisfaction — especially ratings of work-life balance and senior management — and no significant improvement in financial performance or firm value [source: SSRN, 2024]. The authors go further, suggesting the pattern is consistent with managers using mandates to reassert control; that interpretation is theirs and is not itself a measured motive, but the two measured findings — satisfaction down, performance flat — stand on their own.

A crucial caution lives inside that result. RTO being correlated with weaker sentiment, or with any later business outcome, is not proof that it caused it; observational studies can only control for so much. That is exactly why the Bloom randomized trial matters so much as a counterweight, and why the strongest reading combines them rather than leaning on either alone.

There is also a talent dimension that should worry even a mandate's supporters. Using roughly 260 million résumés and a statistical control method, researchers tracked what happened to seniority at Microsoft, SpaceX, and Apple after their return-to-office pushes. They found mandates associated with a measurable drop in average employee tenure — long-tenured, senior staff left at elevated rates, frequently for large direct competitors [source: University of Chicago, 2024]. (Notably, that study found no gender difference in who left; the disproportionate effect on women shows up in the retention and mentoring research, not here.) Losing your most experienced people to rivals is a strange way to strengthen a culture — and it is the risk that independent evidence, as opposed to company assurances, keeps flagging.

The empty-desk backdrop

None of this is happening in a vacuum of real estate. US office vacancy hit a record 20.7% in the second quarter of 2025 — the sixth straight quarterly record — with some markets far worse, such as San Francisco at roughly 27.7% [source: Moody's Analytics, 2025]. Vast, half-empty buildings and the loans behind them create a financial incentive to fill desks that has nothing to do with productivity studies. It would be unfair to claim that any specific company mandated five days to prop up a lease; there is no direct evidence of that. But the backdrop is real, and it explains why the office debate is entangled with balance sheets, downtown tax bases, and commercial mortgages, not just management theory.

Where the equilibrium is settling

Step back from the loudest mandates and a quieter convergence is visible. Global surveys of college-educated workers show working from home has stabilized — at roughly 1.2 to 1.3 days a week, about a quarter of all paid workdays, some 3.5 times the pre-pandemic level, with the post-2022 retreat from remote largely bottomed out [source: PNAS, 2025]. Within hybrid, attendance is clustering on three days in the office, with fewer people at the one-day or four-day extremes; by 2025 roughly two-thirds of companies with formal policies required at least three in-office days [source: Gallup, 2025]. Gallup also finds hybrid employees report the highest engagement, at 35%, versus 33% for fully remote and 27% for fully on-site — an association worth noting, though not proof that the schedule itself causes the engagement [source: Gallup, 2025].

Read together, the hard five-day mandates and the underlying data are pulling in different directions. The mandates push toward the pre-pandemic office; the revealed behavior of workers and firms keeps settling near a middle of about three days. That middle is where the strongest evidence — a neutral-to-positive productivity effect, better retention, and the mentoring value of some in-person time — happens to sit as well.

What to watch

The debate is far from closed, and a few things will tell us where it lands. First, attrition data from the 2025 mandate cohort: if Amazon, JPMorgan, and federal agencies quietly lose disproportionate numbers of senior and hard-to-replace staff, the brain-drain finding gets stronger; if they do not, the mandates look vindicated. Second, whether any large employer publishes measured productivity results — not testimonials — from its five-day policy; so far the case for mandates rests on conviction, while the case for hybrid rests on a randomized trial. Third, the labor market's balance of power: mandates are easier to enforce when employers hold the cards, so watch whether tighter hiring hardens the five-day line or a competitive market pushes it back toward three.

The honest posture is neither "everyone back to the office" nor "the office is obsolete." The measured evidence supports something more specific and less dramatic: hybrid, at around three days, tends to keep productivity intact, retain more people, and preserve some of the in-person mentoring that offices genuinely provide — while blanket five-day mandates carry real retention costs and, so far, no measured performance payoff. Watch what companies measure, not just what they mandate.

Charts

Managers' average view of hybrid work's effect on productivity, before and after the trial

Managers' average view of hybrid work's effect on productivity, before and after the trialBefore the trial -2.6%, After the trial 1%-2.6%Before the trial1%After the trial
Trip.com managers' average estimate of how hybrid work affects productivity, measured before and after the six-month randomised trial. This tracks what managers believed, not what the trial measured about output.Nature (2024) (opens in a new tab)

Change in quit rates under hybrid work, by group (percentage points)

Change in quit rates under hybrid work, by group (percentage points)Women -5.01pp, Non-managers -3.26pp, Long commuters -3.11pp-5.01ppWomen-3.26ppNon-managers-3.11ppLong commuters
Randomised trial of 1,612 employees at Trip.com, 2021-2022. Negative values mean fewer resignations under two work-from-home days a week. The study found no significant effect for managers or for men.Nature (2024) (opens in a new tab)

Employee engagement by work arrangement

Employee engagement by work arrangementHybrid 35%, Fully remote 33%, Fully on-site 27%35%Hybrid33%Fully remote27%Fully on-site
Gallup survey figures for 2025. This is an association between arrangement and engagement, not evidence that moving people on-site lowers engagement — job type and worker choice differ across the three groups.Gallup (2025) (opens in a new tab)

US office vacancy rate

US office vacancy rateA year earlier 20.1%, 2025 Q2 20.7%20.1%A year earlier20.7%2025 Q2
Moody's Analytics data reported for the second quarter of 2025 — a record high for the sixth consecutive quarter. The vacancy backdrop is context for the mandate wave, not proof of any firm's motive.Bisnow, citing Moody's Analytics (opens in a new tab)

Timeline

  1. Randomised trial of hybrid work runs at Trip.com with 1,612 employees — the strongest causal test in this debate.

    Nature (opens in a new tab)
  2. NBER working paper finds engineers in the same building as teammates receive about 22% more feedback on their code, trading short-term output for long-run mentoring.

    NBER working paper #31880 (opens in a new tab)
  3. Nature publishes the hybrid trial: quit rates fall by about a third with no significant effect on performance grades, promotions or lines of code.

    Nature (opens in a new tab)
  4. Ding and Ma study S&P 500 return-to-office mandates and find sharp declines in job satisfaction with no significant improvement in financial performance or firm value.

    SSRN (opens in a new tab)
  5. University of Chicago Harris researchers analyse roughly 260 million resumes and link mandates at Microsoft, SpaceX and Apple to reduced average tenure and seniority.

    Harris School of Public Policy (opens in a new tab)
  6. Amazon CEO Andy Jassy tells corporate staff to return five days a week, ending the three-day hybrid in place since 2023.

    Effective: 2025-01-02

    About Amazon (opens in a new tab)
  7. AT&T moves corporate staff to five days on-site from January 2025 and Sweetgreen raises non-frontline staff to four days.

    Boston Globe (opens in a new tab)
  8. Amazon's five-day mandate takes effect for corporate employees.

    About Amazon (opens in a new tab)
  9. JPMorgan Chase memo orders a companywide return to five days a week, with at least 30 days' notice; more than half of staff were already in the office full-time.

    Effective: 2025-03

    U.S. News / Reuters (opens in a new tab)
  10. A presidential memorandum directs federal agency heads to terminate remote work arrangements and require full-time in-person work, with exemptions; published in the Federal Register on 2025-01-28.

    Federal Register (opens in a new tab)
  11. Dell requires employees living within about an hour of an office to be on-site five days a week; its sales staff had already gone full-time in September 2024.

    Effective: 2025-03

    Boston Globe (opens in a new tab)
  12. US office vacancy hits a record 20.7%, the sixth consecutive record quarter.

    Bisnow, citing Moody's Analytics (opens in a new tab)
  13. PNAS research finds working from home has stabilised at roughly a quarter of paid workdays, several times the 2019 level, with the retreat largely bottomed out.

    PNAS (opens in a new tab)
  14. Gallup reports hybrid attendance clustering at three days on-site, with roughly two-thirds of companies that have formal policies requiring at least three in-office days.

    Gallup (opens in a new tab)

Analysis

The trial tested hybrid, not fully remote

The Nature study gave employees two work-from-home days a week and kept the rest on-site. It is the strongest evidence that a middle position costs nothing measurable in performance — and it says nothing about fully remote teams, which almost no mandate is actually arguing about.

Managers changed their minds when they saw data

Before the trial, Trip.com managers estimated hybrid work cut productivity by 2.6% on average; afterwards, the same managers put it at plus 1.0%. The number that moved was a belief, not an output measure — which is precisely why the debate is so hard to settle by assertion.

Mandates lowered satisfaction without lifting performance

The S&P 500 study found sharp declines in job satisfaction, especially on work-life balance and ratings of senior management, alongside no significant improvement in financial performance or firm value. Absence of a measured gain is not proof that mandates destroy value, but it undercuts the strongest version of the business case.

The people who leave are the ones hardest to replace

The resume study links mandates to lower average tenure and seniority, with long-tenured staff departing at higher rates and often landing at large direct competitors. That is a compositional cost that shows up in institutional memory long before it shows up in a quarterly number.

Only one of these studies can claim causation

The Trip.com trial randomised who worked hybrid, so its comparison is clean. The mandate and resume studies are quasi-experimental, and the Gallup engagement gap is a plain association. Reading all four as though they carry the same evidential weight is the most common mistake in this argument.

Proximity has a real, measured benefit — and a price

The NBER paper found engineers sitting near teammates got substantially more feedback on their code, but that this came at the expense of short-term output, with the tradeoff sharper for women. It is the best empirical case for the office, and it is a case about mentoring, not about hours logged at a desk.

The vacancy backdrop is an incentive, not an explanation

US office vacancy set records for six straight quarters, reaching 20.7% in the second quarter of 2025. That is a real financial pressure on landlords and lease-holding firms, but no evidence in this record shows any specific employer issued a mandate to protect a lease.

The equilibrium is quieter than the headlines

Survey data put working from home at roughly a quarter of paid workdays in 2025, with hybrid attendance clustering at three on-site days. The mandate wave is loud at the top of the market while the broad distribution has largely stopped moving.

Comparison

The 2025 mandate wave: what was ordered, and when
OrganisationDays on-site requiredEffectiveNotes
Amazon5January 2, 2025Announced September 16, 2024; ends the three-day hybrid in place since 2023
AT&T5January 2025Corporate staff
JPMorgan Chase5March 2025Memo in January 2025; at least 30 days' notice; more than half already full-time
Dell5March 2025Applies to staff living within about an hour of an office; sales already full-time since September 30, 2024
Sweetgreen4Announced December 2024Non-frontline staff
US federal agenciesFull-time in personMemorandum of January 20, 2025Exemptions permitted; only about 10% of federal workers were fully remote
What each study can and cannot prove
StudyMethodWhat it measuresWhat it cannot show
Trip.com trial (Nature, 2024)Randomised controlled trial, 1,612 employeesCausal effect of two work-from-home days on quits and performanceEffects of fully remote work, or of firms outside this setting
NBER proximity paper (2023)Quasi-experiment at one Fortune 500 firmFeedback volume and mentoring versus short-term coding outputWhether the tradeoff holds in non-engineering work
Ding and Ma (2024)Difference-in-differences on Glassdoor reviews of S&P 500 firmsChange in job satisfaction and in financial performance after mandatesManagers' actual motives, which the authors interpret rather than measure
Chicago resume study (2024)Distributional synthetic controls, about 260 million resumesShifts in average tenure and seniority after mandates at three firmsWhy individuals left, or the effect on output
Gallup (2025)SurveyAttendance patterns and engagement by arrangementAny causal link between arrangement and engagement
PNAS (2025)Survey seriesWhere working-from-home levels have settledWhat any single employer's policy will do
Moody's AnalyticsCommercial real estate market dataOffice vacancy ratesWhether vacancy influenced any employer's decision
Employer claim versus measured evidence
The claimWho says itWhat the evidence measures
Being together makes learning, brainstorming and culture easier and more effectiveAmazon (company statement)The NBER paper measured more code feedback for co-located engineers — supporting mentoring, at the cost of short-term output
Being together greatly enhances mentoring, learning, brainstorming and getting things doneJPMorgan Chase (memo)The randomised trial found no significant effect of hybrid work on performance grades, promotions or lines of code
Full-time on-site work will strengthen the businessMandating employers generallyThe S&P 500 study found lower satisfaction and no significant improvement in financial performance or firm value
Mandates are how you keep a strong team togetherMandating employers generallyThe resume study links mandates to shorter average tenure and senior departures, often to direct competitors

Process

  1. Announcement

    A memo or executive statement sets out the rationale — culture, mentoring, collaboration. At this stage it is a claim, not a result.

  2. Notice period

    Employees get lead time to rearrange commutes and care: JPMorgan Chase promised at least 30 days.

  3. Effective date

    Amazon's mandate began January 2, 2025; JPMorgan Chase and Dell landed in March 2025.

  4. Compliance and exemptions

    Rules are applied unevenly by distance or role — Dell's applies within about an hour of an office; the federal memorandum allows exemptions.

  5. Measured outcomes

    Attrition, seniority and satisfaction show up in independent research months later. No mandating employer has published measured productivity results.

Sources

  1. Amazon — Update from Amazon CEO Andy Jassy on return-to-office plans and manager team ratio (2024-09-16).View source (opens in a new tab)
  2. Reuters — JPMorgan asks staff to return to office five days a week from March, memo shows (2025-01-10).View source (opens in a new tab)
  3. The Boston Globe — Dell issues five-day return-to-office mandate for employees (2025-02-06).View source (opens in a new tab)
  4. The Boston Globe — AT&T, Sweetgreen demand more days in office, joining Amazon (2024-12-18).View source (opens in a new tab)
  5. Federal Register — Return to In-Person Work (presidential memorandum of Jan 20, 2025; published 2025-01-28).View source (opens in a new tab)
  6. Nature — Bloom, Han & Liang, Hybrid working from home improves retention without damaging performance (2024).View source (opens in a new tab)
  7. NBER — Emanuel, Harrington & Pallais, The Power of Proximity to Coworkers: Training for Tomorrow or Productivity Today? (working paper 31880, 2023).View source (opens in a new tab)
  8. SSRN — Ding & Ma (University of Pittsburgh), Return-to-Office Mandates (2024).View source (opens in a new tab)
  9. University of Chicago (Becker Friedman Institute) — Van Dijcke, Gunsilius & Wright, Return to Office and the Tenure Distribution (2024).View source (opens in a new tab)
  10. PNAS — Aksoy, Barrero, Bloom, Davis et al., The global persistence of work from home (2025).View source (opens in a new tab)
  11. Gallup — Hybrid Work in Retreat? Barely. (2025).View source (opens in a new tab)
  12. Moody's Analytics — U.S. office vacancy hits another record high (2025).View source (opens in a new tab)

Tags

  • #return-to-office
  • #hybrid-work
  • #remote-work
  • #workplace-productivity
  • #future-of-work