The federal wage survey carries no remote-work flag — it records where a job is reported, not where the person sits. Estimates of remote share come from household surveys asking about telework, and from private trackers counting job postings. Those answer different questions and produce different numbers, so any single remote-share figure needs its source and definition attached before it means anything.
Why the wage data cannot answer this
No official statistic reports what share of a given occupation is remote. The private trackers that try to measure it disagree with each other, because they are counting genuinely different things. The disagreement is a definitional one rather than a factual one.
Occupational employment is reported by establishment location. A fully remote worker is generally counted where the employing establishment sits rather than where they actually live and work. Nothing in the survey asks where the person physically sat.
So the geography in wage surveys is employer geography rather than worker geography, and it always was. That limit grew quietly with remote work and it constrains every map-shaped claim about pay, including the ones on this site. Concentration describes where work is bought, not done.
Three different questions, three different numbers
The first question is whether people teleworked, which household surveys ask directly. That captures partial and occasional remote work, so it produces the largest figures and the vaguest ones, since one day a fortnight counts the same as five. The result is a large number covering very different weeks.
The second is whether postings say remote, which private trackers measure by scraping listings. That records what employers advertise, which moves faster than what they practice and responds to how hard they are finding it to hire. Advertised flexibility is a recruiting lever in its own right.
The third is whether the work could be done remotely at all, which comes from task-based classification of occupations. That is a capability estimate rather than an observation, and it produces numbers that are true while describing nobody’s actual week. Headlines quote one of the three and describe another constantly.
Why postings overstate it
A remote listing attracts several times the applicants of an on-site one for the same role. That is a strong incentive to advertise flexibility the role does not fully carry, and it costs the employer nothing at the posting stage. The correction happens later, in conversation with a manager.
Posting counts also double-count syndication, so one role appearing on four boards looks like four separate remote opportunities. Nothing in the count corrects for that. Syndication inflates the visible total without adding roles.
And remote roles are disproportionately posted publicly rather than filled internally. That inflates their share of what you can see without inflating their share of what actually exists, which is a subtler distortion and a larger one. Internal hiring is invisible to anybody scraping listings.
What is reasonably established
Remote work concentrates where the output is information and is near-absent where physical presence is required, which is most work. That much is consistent across every source regardless of method. Where the sources disagree is on magnitude rather than direction.
The occupations with high remote shares are largely the same ones with high exposure to language models. Both follow from the same underlying property, which is that the work is text, code and analysis rather than objects and people. The same property drives both distributions at once.
It is also stratified by seniority within the same occupation. The remote share of senior roles is typically higher than the junior share, because the case for being in the room is strongest when somebody is still learning the work. Junior roles are the last to go remote and the first to return.
The question that actually decides your situation
Ask the employer how many people on the specific team work remotely, and how often the rest come in. One honest answer about one team beats every national percentage available. The specific answer is also far easier to obtain.
You are not joining a national percentage and you are not joining an occupation. You are joining a team with a particular manager and a particular set of habits. Those habits vary enormously inside a single company.
Then ask what happens when policy changes, because it has changed repeatedly at most large employers since 2020. What matters is whether your arrangement is a written term of employment or merely a current practice, and those differ enormously when somebody new takes over. A written term survives a change of management and a practice does not.
The pay dimension
Whether a remote role is priced by the employer’s location or by yours changes the number substantially. Both policies are common and neither is announced unless you ask. The offer letter frequently says nothing either way.
Establish which applies before accepting anything, and establish separately what happens if you move. Those are two different questions and employers sometimes answer them inconsistently. An inconsistent answer is itself worth noticing carefully.
There is a genuine trade underneath. Location-indexed pay tends to be lower for you and more durable for the employer, while a single national rate is better for anybody outside the expensive metros and more likely to be revisited when budgets tighten. Neither arrangement is permanent and both are worth confirming.
Why this matters for reading pay data at all
If a growing share of an occupation works somewhere other than where it is counted, then metro wage figures for that occupation describe the employers rather than the workforce. For heavily remote occupations that gap widens every year. The measure was designed before the gap existed at all.
The practical consequence is that a metro median for a highly remote occupation is best read as what employers headquartered there pay. That is still useful, because it is the number you would be negotiating against. That is the practical use of a figure with a known limit.
What it is not is a statement about local living standards in that metro. For occupations that must be done in person the two meanings coincide and the figure carries both, which is worth knowing before drawing conclusions about a place from a wage table. The same number means different things for different occupations.
How to read any remote-share figure you encounter
Ask three things of any number before using it. Which of the three questions it answers, what period it covers, and whether it counts occasional remote work the same as fully remote. Those three questions produce very different answers indeed.
A figure missing any of those is not usable for a decision about your own job. It may still be interesting and it is not evidence about anything specific. Interesting and usable are genuinely different standards here.
That test disqualifies most of what circulates, which is the point rather than a problem. The honest version of this subject is that the national picture is genuinely unclear and your own team’s practice is knowable in a single conversation. Spend your effort on the knowable half of that.
Common questions
Is there an official remote-work share by occupation?
No. The federal wage survey carries no remote flag and records jobs by establishment location rather than where the worker sits.
Why do remote estimates differ so much?
They measure different things — whether people teleworked, whether postings say remote, or whether the work could be done remotely.
Do job postings overstate remote work?
Generally yes. Remote postings attract far more applicants, which rewards advertising flexibility the role may not fully carry.
Which occupations are most remote?
Those whose output is information — largely the same ones with high language-model exposure, for the same underlying reason.
What does this limit on the data mean?
That geography in this dataset is employer geography rather than worker geography, which constrains every geographic claim.
Is there an official remote-share figure by occupation?
No. Wage surveys carry no remote flag and record jobs by establishment location, so the geography in them is employer geography rather than worker geography.
Why do remote estimates disagree so much?
They answer different questions — whether people teleworked, whether postings say remote, or whether the work could be done remotely. Only the first is an observation of behavior.