The warning is common enough to be furniture. Take the remote job and you will be the one nobody remembers at promotion time.
It has been measured. Twice by randomised experiment, once by watching a firm's personnel records through a shock that made everyone remote at the same moment. The three results do not agree. The most conspicuous difference between the settings is how much of the week the workers were away, though no study varies that on its own, so the reading is a hypothesis rather than a measurement.
The first experiment found a promotion penalty
In late 2010, Ctrip, a 16,000-employee NASDAQ-listed Chinese travel agency, asked 996 people in the airfare and hotel departments of its Shanghai call centre whether they would like to work from home four days a week, with the fifth day in the office as usual. Just over half volunteered; 249 of them qualified. Those with even-numbered birthdates worked from home, those with odd-numbered birthdates stayed in the office, and the arrangement ran for nine months.1
The productivity results are the famous part. Performance rose 13 per cent, of which about 9 points came from working more minutes per shift, mostly fewer breaks and sick days, and 4 points from handling more calls per minute, which workers put down to the quieter room. Two separate quality metrics showed no deterioration. Attrition fell by half.1
Promotions were tracked for 22 months, to the end of September 2012. Forty of the 249 were promoted to team leader or moved to a more advanced function, a promotion rate of 15.9 per cent: 17 from the home-working group and 23 from the office group. Compared head to head, that difference is not statistically significant, and the paper says so: "working from home has no overall impact on promotion".1
The finding appears once performance is held constant. Home workers had performed better, and performance strongly predicted promotion, so the raw comparison is a better performer being promoted at the same rate as a worse one. Once performance during the experiment was controlled for, working from home carried "a negative and weakly significant impact on promotion", which the authors summarise as reducing rates of promotion by about 50 per cent.1
Their explanation is a phrase everyone already knows. Home-based employees are "out of sight, out of mind", so supervisors noticed their performance less. They offer one alternative: that home workers had fewer chances to build the interpersonal skills a team-leader job needs.1
What happened next is the part worth remembering. Ctrip rolled the option out to the whole firm and let the experimental employees choose again. Two thirds of the office group, all of whom had volunteered to work from home ten months earlier, decided to stay in the office, "citing concerns over the loneliness of home working and lower rates of promotion".1 The people in the building had drawn the same conclusion the regression did, and acted on it.
The same company ran it again and got a different answer
Ctrip is now Trip.com, and the Nature paper reporting what happened next describes the 2010 study as the same firm's own earlier work.2 In August 2021 it ran a second randomised trial, this time on 1,612 graduate employees in its Airfare and IT divisions in Shanghai: 395 managers and 1,217 non-managers, working in engineering, marketing and finance. Employees born on odd-numbered days were given the option of working from home on Wednesday and Friday and coming in the other three days. Employees born on even-numbered days came in five days a week. The results were published in Nature in June 2024.2
Attrition fell by a third, from 7.2 per cent in the control group to 4.8 per cent. The reduction was 54 per cent among women and 52 per cent among people with above-median commutes, and there was no significant effect among managers. Job satisfaction rose. The 395 managers, surveyed before and after, moved from expecting hybrid working to cost 2.6 per cent of productivity to reckoning it had added 1.0 per cent.2
On careers the paper is blunt: "We found no evidence of a difference in promotion rates over the next two years overall, or for any major employee subgroup."2 Performance grades across four consecutive half-year review rounds showed no difference either.
Three things stop that being the end of the argument.
The treatment was an option rather than an obligation, and take-up ran at about 55 per cent among volunteers and 40 per cent among everyone else, which, as the paper notes, means both groups "tended to WFH only one day, typically Friday, each week".2 The experiment measured the effect of being offered two days at home in a firm where most people took one.
"No evidence of a difference" is also not the same as evidence of no difference, and the authors are careful about this. Their equivalence tests could rule out a promotion gap larger than two percentage points in two of the four half-year periods, and could not in the other two.2
And they say plainly where their result stops: "We are not sure whether the results would extend to more remote settings such as one day a week (or less) in the office, owing to potential challenges around training, innovating and culture in fully remote settings."2 They cite the fully remote evidence themselves.
Then everyone went remote, and the gap closed downwards
That evidence comes from a US Fortune 500 firm that ran remote and on-site versions of the same call-centre job, which let Natalia Emanuel and Emma Harrington compare people doing identical work under different arrangements from January 2019 to October 2021.3
Before the pandemic the promotion difference was enormous. A year after being hired, 44.0 per cent of on-site hires had been promoted, against 20.9 per cent of remote hires. Remote and on-site staff sat on different teams under different managers, so they were not competing with each other for the same openings.3
Then the offices closed and every worker became remote. The gap in monthly promotion rates between the two groups disappeared, and it matters which way round that happened. It closed because the promotion rate of the people who had been in the office fell, not because the remote group caught up. That fall is the study's estimate of what remote work does to a career: 3.6 percentage points a month, against an on-site base of 6.1 per cent, or 58.7 per cent of it. Which is close to what the Ctrip trial found a decade earlier and half a world away.3
That disposes of the comfortable version of this, in which the problem is being the odd one out and the fix is joining a company where everybody is remote. In the one firm that ran the experiment by accident, nobody was the odd one out afterwards and promotions fell anyway.
The same data show what the office was supplying. When the buildings were open, on-site hires spent more time in training sessions and more time in one-to-ones with their manager "planning their short-term path to promotion over the next 30, 60, and 90 days". Remote work cut training time by 19.1 minutes a month, or 26.3 per cent, and manager one-to-one time by 10.2 minutes, or 34.1 per cent. Both advantages vanished when the offices shut.3
There is a hopeful footnote. Those promotion figures count everyone, including people who left. Among workers who stayed at the firm, remote workers caught up with their on-site colleagues at around 15 months.3
What proximity buys, and who pays for it
Emanuel and Harrington, with Amanda Pallais, went looking for the mechanism among software engineers at a Fortune 500 online retailer between 2019 and 2024. Some teams sat in one building, some were split across two headquarters buildings a ten-minute walk apart, and some were spread across the country. The 2020 closures flattened those differences and the return-to-office mandates of 2022 and 2023 restored them.4
Engineers on single-building teams received 23.9 per cent more comments on their code than engineers split across two buildings. When the offices closed, that advantage narrowed by 18.3 per cent, and the comments that disappeared were disproportionately the ones a classifier rated helpful, actionable, impactful and well reasoned. A ten-minute walk across a campus cost about as much feedback as being several states away. One remote hire was enough to change a team: when a new joiner turned a co-located team into a distributed one, feedback dropped between the people who were still sitting together.4
The benefit was not spread evenly. Less-tenured and younger engineers gained most, and paid most when it went. Senior engineers, who supply the feedback, wrote less code when they sat near their teams — mentoring is unpaid work that comes out of somebody's day.4
This measures code review at one company, and feedback is not promotion. But it names what the two call-centre studies could only infer: what proximity delivers is the attention of people who could teach you something, and that attention turns out to be fragile in a way that distance in miles does not capture.
If you are early in your career
The same paper reports something outside its firm, and the change in evidence quality should change how much weight you give it. Comparing US occupations that can be done remotely with ones that cannot, the authors find that between 2017–2019 and 2022–2024, young graduates in remotable occupations saw a 0.65 percentage point larger rise in unemployment than older graduates in the same occupations or young graduates in work that cannot be done remotely.4 The authors head off the obvious rival explanation, noting that the gap opened before generative AI spread and survives adjusting for how exposed an occupation is to it. It is still a pattern in national statistics rather than an experiment, and it is in a working paper that has not been peer reviewed.
The firm-level finding beside it is easier to trust and points the same way: the company shifted towards hiring older workers while its offices were closed, and back towards younger hires once they reopened.4 If a job is one where you would be learning it, that is worth knowing before you take the fully remote version of it.
What to ask before you accept
The advert's location line is not yet a term of your contract, so the interview is the moment to find out how this particular arrangement works.
Ask how many days a week you would actually be at home, and pay more attention to that number than to the job's label. The trial that found no evidence of a penalty was one or two days a week; the two that found one were four days and full time. That is a pattern across three settings rather than a measured contrast, so treat it as the best available guess, not a dial.
Ask how the team is spread out, and do not assume that everyone being remote settles it. One distant hire was enough to cut the feedback flowing between colleagues who were still sitting together, and the firm whose whole workforce went home saw promotions fall rather than even out.
Ask how much one-to-one time with your manager is scheduled, and what happens in it. The measured mechanism was a third less manager time and a quarter less training, not fewer hours worked.
Ask who has been promoted out of this team in the last two years, and where they worked. It is a fair question, it is answerable, and the answer is more informative than any policy document.
Three of the four studies here measure promotion: two in call centres, and one among engineering, marketing and finance staff at a Chinese travel company. The fourth measures code-review feedback among software engineers, which is a mechanism rather than an outcome. All three look inside a single employer, so none of them says anything about the step up you get by moving to a different one. None of them varies the days at home while holding everything else steady either: the two Chinese trials are eleven years and two different workforces apart, and the third is another country, another employer and a natural experiment rather than a randomised one. And the trial closest to an ordinary hybrid week found no evidence of a penalty, alongside a third fewer people quitting, so nothing here argues for going back to five days in the office.
What the three promotion studies agree on is narrower. Where there is a cost, it is paid in contact with the person who will be asked whether you are ready. You can ask about that now, instead of finding out at your first review.
