BLS OEWS Wage Data: What It Proves in a Reasonable Compensation Study, and What It Does Not

Most reasonable compensation files contain at least one number that came from the Bureau of Labor Statistics. It sits in the workpapers as a wage figure with an occupation title beside it, and in a lot of files that is the entire citation. Thin record. The figure carries structure nobody wrote down: a survey, a population of workers, an occupation code, a geography, a reference period, a point in a distribution. Change any one and the number changes.

This page is about the evidence source itself. What the Occupational Employment and Wage Statistics program measures, how its codes, areas, periods and percentiles are built, and where the seams open when one of its figures carries a conclusion about a single shareholder-employee. The standard that figure gets measured against is a separate subject: the IRS guidance trail, what a reasonable salary means, and the officer rules cover it.

Everything below was read on the BLS site on September 20, 2026, and describes the May 2025 estimate set.

What OEWS measures, and who is counted in it

OEWS is an employer survey of occupational employment and wage rates for wage and salary workers in nonfarm establishments. BLS draws the sample from state unemployment insurance files, stratified by area, industry and size, with larger employers more likely to be selected. The full sample is about 1.1 million establishments.

The scope line that matters most for an S-corporation file sits in the definition of employment. OEWS counts salaried officers, executives and staff members of incorporated firms. It does not count the self-employed, owners and partners in unincorporated firms, household workers, or unpaid family workers.

So a shareholder-employee is inside the surveyed population in principle, as a salaried officer of an incorporated firm. But they enter only through an employer's response about an occupation, never as an identified owner, and no field separates owner pay from employee pay. That is the first seam, and it is structural. OEWS is evidence about a labor market, which a study has to connect to a person.

The wage definition narrows things further. OEWS wages are straight-time gross pay, exclusive of premium pay: base rate, cost-of-living allowances, guaranteed pay, hazardous-duty pay, incentive pay including commissions, and tips. Overtime, severance, shift differentials, nonproduction bonuses, employer cost of benefits and tuition reimbursement are out. Annual figures for establishments reporting in wage intervals come from multiplying the hourly rate by a 2,080-hour work year.

The occupation code classifies the work, and an owner usually does several jobs

OEWS classifies occupations under the 2018 Standard Occupational Classification, which sorts all workers into 867 detailed occupations, combined into 459 broad occupations, then 98 minor groups and 23 major groups. The May 2025 estimates carry about 830 occupational categories across 22 of the 23 major groups, military occupations excluded, and some are published at the broad level or as OEWS-specific combinations rather than in detail.

Every one of those codes describes a set of job duties. None describes an owner. The person running a small S corporation sells, delivers the service, manages staff, signs contracts and handles the money, and those activities sit in different SOC major groups. Pull one code and report its wage, and the file has answered a question about a job nobody in the company holds. That is why the split matters more than the source. A study that allocates the shareholder's time across several occupations and cites a figure for each leaves a record someone can inspect. A study that picks the closest-sounding single title produces a number that reads the same and cannot be checked.

Arithmetic will not rescue a weak split. BLS does not recommend averaging percentile wages, because a percentile for a combination of occupations typically will not equal the average of the percentiles of its parts. Its published example uses human resources workers, broad occupation 13-1070: for May 2024 the 90th percentile annual wage was $128,180, while the simple average of the three detailed occupations came to $122,280 and the employment-weighted average to $128,292. Neither equals the published figure.

Geography is a published area level, not a neighborhood

OEWS publishes cross-industry estimates for the nation, for states, the District of Columbia, Guam, Puerto Rico and the U.S. Virgin Islands, and for approximately 530 metropolitan statistical areas and nonmetropolitan areas. The May 2025 estimates use the metropolitan definitions in OMB Bulletin 23-01; nonmetropolitan definitions are specific to OEWS and set with the state workforce agencies.

There is no finer level. A metropolitan estimate covers the whole defined area, not the client's town, so geography is a choice among named levels that needs a stated reason.

Three years of panels stand behind one May reference date

OEWS combines six semiannual panels collected over three years, two per year at roughly 186,000 to 189,000 establishments each, contacted in May and November. The May 2025 estimates are built from May 2025, November 2024, May 2024, November 2023, May 2023 and November 2022. Unweighted sampled employment across those panels was 84.7 million, about 55 percent of national employment, at a 66.2 percent response rate by establishment across the 50 states and the District of Columbia.

Wages from the earlier panels are adjusted forward to the current May reference date by a modeling procedure, and the estimates come from a model-based method BLS calls MB3, which models staffing patterns and wages for every establishment in the population it did not observe. BLS says directly that this design limits the usefulness of OEWS data for time series analysis. Estimates are generally released in late March or early April.

Two consequences land in the workpapers. A figure's reference period is not its download date, so the file records both, and a year-over-year change in an OEWS figure is not evidence the market moved.

A percentile is a boundary, not a standard

BLS defines a percentile wage as the value of a wage below which a certain percentage of workers fall. The median is the 50th: half the workers in the occupation earn more, half less. That is the whole meaning.

Nothing in the OEWS product designates any percentile as correct, appropriate, or safe. BLS publishes a distribution and its definitions, not a recommendation. A percentile chosen in a study is a judgment defended from the study's own facts, and a file reporting the 75th percentile with no stated reason has recorded a preference, not evidence.

Two publication limits sit on top of that. OEWS withholds some percentile wages for especially high-paying occupations, publishing only a note that the wage is at or above a cutoff. Others are withheld where they do not meet publication standards, so an occupation can be missing entirely for an area.

What the file has to show

A wage figure is not evidence until the record says how it was chosen. Five things carry it, and every one is a property of the source.

Which occupations, and why that split. Every SOC code used, the share of the shareholder's time assigned to each, and the duties behind the assignment. One code for a multi-role owner is the most common weak point.

Which geography level, and why. National, state, or a named metropolitan or nonmetropolitan area, stated as the level it actually is.

Which release and which reference period. The estimate set, its reference date, and the date accessed, which is how BLS asks that OEWS be cited.

Which percentile, and the reason for it. If the figure was withheld or published only as a cutoff, the file says so instead of substituting a nearby one.

What the figure excludes. Straight-time gross pay carries no benefits, no overtime, no nonproduction bonuses, so comparing it to a shareholder's total package compares two different things.

That record is what makes a conclusion inspectable years later. The workpapers a study leaves behind, what holds up when the file is examined, and the pillar all turn on it. Every figure traces to its source, occupation, geography and period. The Reviewer signs off on the record.

TracePrep is a software product from TracePrep Inc., not a CPA firm. This content is operator experience for educational purposes; it is not tax advice and has not been reviewed by a licensed CPA. Consult a qualified tax professional for guidance on your specific situation. Source citations in TracePrep workpapers support audit defense but do not guarantee IRS outcomes.