Two days a week at home makes an employee 7.7% less likely to be picked for promotion and 7.1% less likely to be picked for a raise — by managers choosing between otherwise identical candidates (Kasperska, Matysiak & Cukrowska-Torzewska, PLOS ONE, 2024). Full-time remote widens it to 10.7% and 9.4%.
Now the part that matters operationally: in the same experiment, the penalty disappeared entirely when most of the team worked from home too.
That is the shape of the remote work promotion penalty as the evidence now describes it. It is not a fixed cost of working from home. It is a variable cost set by how common — and how normal — the arrangement looks inside your company. A new factorial vignette experiment published in Gender, Work & Organization has just replicated that finding in a second country and pushed it further: the penalty tracks how the policy is framed, not how many days anyone is actually absent (Wang & Chung, Gender, Work & Organization, 2026).
Which makes hybrid policy a promotion-calibration problem, not a real-estate one.
What the 2026 Experiment Actually Measured
Senhu Wang (National University of Singapore) and Heejung Chung (King's Business School) ran a factorial vignette study on a national sample of 473 managers, each rating a set of hypothetical employee profiles that varied on gender, parenthood, remote-working pattern, and the surrounding workplace context (King's College London, 2026; Fast Company, 2026).
Managers rated each profile on four dimensions: commitment, productivity, team spirit, and promotion opportunity.
The headline result is uncomfortable and consistent. Remote working produced significantly worse managerial perceptions on all four dimensions, for every gender–parenthood group — mothers, fathers, childless women, childless men. Not one group was exempt. The authors report similar patterns in UK and German data.
A vignette experiment is the right instrument here, and it is worth understanding why. You cannot detect this bias in your own promotion data, because in the real world remote workers differ from office workers in a hundred unobserved ways — role, tenure, manager, ambition, commute. The vignette design holds all of that constant and varies one attribute at a time. What you are seeing is not a correlation. It is the causal weight a manager assigns to the attribute "works from home two days a week," with performance and everything else held fixed.
The trade-off is external validity: managers are rating profiles, not people they know. Read the magnitudes as a measure of the bias managers bring to an ambiguous case, not as a prediction of what happens to a top performer whose work the manager sees daily.
The Remote Work Promotion Penalty Has a Number, and It Is Not Small
The Wang and Chung study tells you the direction and the moderator. The UK conjoint experiment tells you the size.
Kasperska and colleagues put 937 British managers through a forced-choice design — pick one of two candidates for promotion, a raise, or training — varying working mode alongside age, experience, skills, and performance rating. Holding performance constant, hybrid workers came out 7.7% less likely to be chosen for promotion and 7.1% less likely for a salary increase. Full-time remote workers: 10.7% less likely for promotion, 9.4% for salary, and 6.6% less likely to be selected for training (PLOS ONE, 2024).
The training number is the one operators should sit with. A promotion penalty is a single event. A training penalty compounds: fewer development slots this year produce a weaker case next year, and the gap that started as a perception becomes a documented capability difference. Bias that reduces investment doesn't stay a perception problem. It manufactures the evidence that justifies it.
Two experiments, two countries, two designs, same conclusion — with performance explicitly controlled in one of them.
The Penalty Lands on the People You Did Not Expect
Both studies converge on the same counterintuitive distribution: the worst effects fall on fathers and childless workers, not on mothers.
In the UK experiment, mothers working from home experienced no significant promotion or pay penalty at all. Men — fathers and non-fathers — and childless women took the hit. The Singapore experiment reports the same asymmetry: effects more pronounced for non-mother groups, especially fathers.
The mechanism is not generosity toward mothers. It is that managers already discount mothers' commitment at baseline, so remote work adds little new information; for everyone else, it is a fresh signal that gets read as withdrawal. Mothers are not spared the bias. They have already paid it.
This should change how you audit. If your fairness review looks only at gender or only at parental status, you will find nothing and conclude the policy is clean. The exposed groups here are fathers and childless staff of both genders — populations almost no DEI dashboard segments for, and populations that in a 200-FTE company make up most of the promotion pipeline.
The Variable You Control Is Framing, Not Attendance
Here is the finding that turns this from an interesting result into a decision.
In the UK conjoint, when 80% or more of the team worked from home, the promotion and pay penalties were eliminated for every group. Managers who themselves worked from home frequently did not penalize remote workers at all. In the Singapore vignettes, the same logic held: where remote work looked like the normal option for everyone, the stigma weakened toward noise; where it was framed as an accommodation for mothers or parents, it persisted and intensified.
Chung's summary is blunt: "All workers suffer from career penalties when working remotely, and this is especially true when it is primarily framed as a solution for parenting or care" (King's College London, 2026).
Read that against how most mid-market hybrid policy is actually written. The common pattern — a default of in-office, with remote days available on request, approved case by case, often justified internally as flexibility for working parents — is the exact configuration the evidence identifies as maximally penalizing. Every approval marks the employee as an exception. Exceptions are what managers rate down.
The well-intentioned version of the policy is the expensive one. A blanket two-days-remote default for everyone, with no application and no approver, costs the same in square footage and removes the signal entirely.
The Counter-Evidence: Performance Says Hybrid Is Fine
The strongest objection to acting on this is that perception studies measure perception. So look at what happens when someone runs the real experiment on real output.
Bloom, Han and Liang randomized 1,612 employees at Trip.com into hybrid or full-time office work for six months and tracked them for two years. Hybrid workers were as productive as their in-office peers on performance reviews, quit rates fell sharply among non-managers, and — critically — hybrid employees were promoted at the same rate (Bloom, Han & Liang, Nature, 2024).
That is not a contradiction. It is the whole problem stated precisely.
Trip.com ran hybrid as a company-wide default, assigned by randomization, applied to entire teams. Under those conditions the penalty vanished — exactly as the vignette studies predict it should when the arrangement is universal rather than individual. The Nature result is not evidence that proximity bias is imaginary. It is evidence of the one configuration in which it does not bite.
Your company is unlikely to be running that configuration. If remote days in your organization are individually negotiated, unevenly distributed across teams, and concentrated among caregivers, you have built the vignette condition, not the Trip.com condition.
The Objection: "Our Managers Promote on Output"
They believe they do. So did the managers in both experiments — who were shown explicit performance ratings and still moved 7.7 percentage points on working mode alone.
Two clarifications worth holding onto. First, none of this requires anyone to be acting in bad faith; the effect is a default inference about commitment drawn from reduced visibility, and it operates fastest precisely where the evidence is thinnest. Second, the effect is largest for ambiguous mid-tier cases — the bulk of any promotion slate — and smallest for people whose output is unmistakable. Your top 5% are probably fine. Your bench is not.
And managers who work remotely themselves show no penalty at all. If your leadership team is in the office five days while the organization is hybrid, you have a structural calibration gap, not an attitude problem.
What to Do Before the Next Promotion Cycle
Three moves, all executable this quarter.
1. Read your policy for exception-generation. Find every clause requiring an employee to request, justify, or get approval for remote days. Each one converts a working arrangement into a personal accommodation, and the research says accommodations are what get rated down. Convert to a stated default that applies to whole teams. Same cost, no signal.
2. Strip the caregiver framing from the language. If your policy, intranet page, or manager talking points justify flexibility as support for working parents, rewrite them. The intent is good and the effect is measurably backwards — it concentrates the stigma rather than distributing it, and it penalizes fathers and childless staff hardest.
3. Run the audit segmented correctly. Compare promotion rates, raise rates, and training allocations for remote-heavy staff against in-office peers at matched performance ratings — segmented by parental status and gender, with fathers and childless employees as their own cells. Training allocation is the leading indicator; it moves a year before promotions do.
The Question for Your Next Calibration Meeting
The evidence does not say the remote work promotion penalty is the price of working from home. It says the penalty is the price of working from home when it looks like a favour.
That is a policy-drafting decision, made once, by someone in your organization — probably without knowing it was a compensation decision. Before the next calibration meeting, pull up your hybrid policy and ask one question: does an employee have to ask permission? If the answer is yes, you are not running a flexibility policy. You are running a promotion filter, and you are paying for it in the exact population you can least afford to lose.
