What BLS Wage Data Is, and What It Leaves Out

The three it does not announce

You find the median for your occupation in your metro, and it is lower than every figure you have seen elsewhere. Before deciding the market has moved, it is worth knowing what that number was built to measure. The federal wage survey is the most defensible salary source available to anybody, and it is also silent about three things that matter enormously depending on who you are. None of the three is disclosed on the page where you read the figure.

The first is the self-employed, who are not in it at all. The survey samples establishments and asks what they pay employees, so anybody without an employer is absent by construction. Owners and partners of unincorporated firms are excluded, as are agricultural workers on small farms and some household workers. In an occupation where a large share of people work for themselves, the published figure describes only the employed portion, and that portion may be systematically different from the rest.

The second is employer benefits. The survey reports straight-time gross wages and salaries, which means no health premium, no retirement contribution, no paid leave, and no annual bonus or stock. For a job where the employer covers most of a family health plan and matches retirement generously, the published wage understates the package by a considerable margin. Comparing two occupations on wages alone quietly assumes their benefits are similar, and they often are not.

The third is the calendar. The figures describe a reference period roughly a year before you read them, because a survey of that size takes time to collect, clean and publish. In a stable market that hardly matters, and in a fast-moving one it matters a great deal. The number is accurate about a moment that has already passed.

Four more it does not contain

Years of experience are not collected, not published and not derivable from what is published. The tenth percentile mixes genuine beginners with part-time-equivalent workers and low-paying regions, and nothing in the data separates them. Any site presenting a clean experience curve built on this source is inferring rather than reporting. That inference may be reasonable, and it is still an inference.

Sex, race and ethnicity are also absent, because an establishment survey asks employers about positions rather than about people. Pay gap analysis therefore comes from a different, household-based survey with its own methodology and its own sample. Mixing the two sources inside one argument is a common and avoidable error. Remote work is missing in the same way, since jobs are counted where the establishment sits rather than where the worker does.

Employer identity and size are the fourth gap, and they close off a question people ask constantly. You cannot use this data to compare what large employers pay against small ones, or what one named company pays against another. That question has real answers elsewhere, in self-reported datasets and advertised ranges. It simply is not answerable here.

Why the omissions are deliberate

Every one of those gaps is a design decision rather than an oversight. Employers can report what they pay for a defined position accurately and cheaply, which is why the survey asks that and nothing else. Asking them for worker demographics, remote status or individual tenure would produce worse data at higher cost. Response rates would fall, and the thing that makes the survey valuable would degrade.

So the omissions are the price of the survey’s greatest strength. It has an enormous sample and consistent definitions precisely because it asks one narrow question well. A source that tried to answer everything would answer nothing with this much confidence. Knowing which question it answers is most of what you need to use it properly.

What that means when you use it

It answers what an occupation pays in a place, with real authority and a sample nothing else comes close to. It cannot answer what somebody with eight years of experience should earn, or what a particular employer pays, or what your total package is worth. Those are three different questions with three different sources behind them. Asking this one for an answer it does not hold is how people end up quoting a figure that does not survive scrutiny.

Experience-based figures come from private compensation surveys, which have smaller samples, proprietary methods and paying customers. That is not disqualifying and it is a genuinely different kind of source. The discipline is naming which source answered which part of your question. A figure assembled from three places that admits to none of them is the least trustworthy kind there is.

Suppression, and why some cells are blank

Where publishing a figure would identify an individual employer, or where the estimate does not meet quality standards, it is withheld. That is why some occupations show employment but no wage, and why a metro you know has these jobs sometimes shows nothing at all. The rule exists to protect the employers who supply the data, which is what keeps them supplying it. Without it the survey would have far fewer respondents and far worse coverage.

A blank is not a zero and it is not an error in the file. It usually means the occupation is small in that area, which answers a question you were probably also asking. If almost nobody in your metro does this work, that is information about whether the place is a market for you. Read the gap rather than skipping past it.

The annualization assumption

Annual figures for hourly occupations assume full-time year-round work at 2,080 hours. For occupations with substantial part-time or seasonal employment, the published annual figure is a full-time equivalent rather than what anybody takes home. An occupation paying $18 an hour shows an annual figure of $37,440 on that assumption. Somebody genuinely working thirty hours a week earns $28,080 doing the same job at the same rate.

That $9,360 difference is not an error in the data and it is not visible in the number either. It inflates apparent annual pay in retail, food service and parts of healthcare relative to lived experience. When somebody compares the published table against their own year and concludes the table is wrong, this is usually why. Check the hourly figure instead, which carries no such assumption.

How to fill the gaps honestly

For experience, use private salary surveys or advertised ranges in your metro, treated openly as weaker evidence. For demographics, use the household survey and cite it as the household survey. For total compensation, use the employer’s own plan documents, because nothing public covers your specific package. Each substitution is defensible as long as you say you made it.

The version that fails is the confident blended number with no stated sources behind it. It falls apart the moment somebody asks where a component came from, and it takes the rest of your case with it. Name the source for each piece and the whole thing survives questioning. That is worth more in a negotiation than a larger number you cannot defend.