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
Occupational employment is reported by establishment location. A fully remote worker is generally counted where the employing establishment sits, not where they live.
So the geography in wage surveys is employer geography, not 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, which is why the concentration figures are described as where the work is bought rather than where it is done.
Three different questions, three different numbers
Did people telework? Household surveys ask directly. This captures partial and occasional remote work, so it produces the largest figures and the vaguest ones — one day a fortnight counts.
Do postings say remote? Private trackers scrape listings. This measures what employers advertise, which moves faster than what they practice and responds to how hard they are finding it to hire.
Could the work be done remotely? Task-based classification of occupations. This is a capability estimate rather than an observation — the same distinction as automation exposure, and it produces numbers that are true and describe nobody’s actual week.
Headlines quote one and describe another constantly. A figure without its definition attached is not usable for a decision about your own job.
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 reason to advertise flexibility the role does not fully carry, and it costs the employer nothing at the posting stage.
Posting counts also double-count syndication: one role appearing on four boards looks like four. And remote roles are disproportionately posted publicly rather than filled internally, which inflates their share of what you can see without inflating their share of what exists.
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. The occupations with high remote shares are largely the same ones with high exposure to language models, because both follow from the work being text, code and analysis.
It is also stratified by seniority within the same occupation. The remote share of senior roles in a field is typically higher than the junior share, because the case for being in the room is strongest when somebody is still learning the work.
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, because you are not joining a national percentage.
Then ask what happens when policy changes, because it has changed repeatedly at most large employers since 2020 and there is no reason to assume it has stopped. What matters is whether your arrangement is a written term of employment or a current practice, and those are very different things when somebody new takes over.
The pay dimension
Whether a remote role is priced by the employer’s location or yours changes the number substantially, and both policies are common. Establish which applies before accepting, and establish what happens if you move.
There is a genuine trade here. Location-indexed pay tends to be lower for you and more durable for the employer; a single national rate is better for anybody outside the expensive metros and more likely to be revisited when budgets tighten.
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 practical consequence: a metro median for a highly remote occupation is best read as what employers headquartered there pay, which is still useful — it is the number you would be negotiating against — but it is not a statement about local living standards. For occupations that must be done in person, the two meanings coincide and the figure carries both.
Worth knowing which kind you are looking at before drawing conclusions about a place from a wage table.
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.