What union wage tables show and do not

Once a year BLS publishes the union membership rate and, beside it, median weekly earnings for union members and for nonunion workers. The union versus nonunion earnings page carries that series. The two medians are often read as a measure of what joining a union is worth. BLS itself says they are not, and this page explains why.

Where the data come from

The figures are from the Current Population Survey, the monthly household survey of about 60,000 households. Both the union questions and the earnings questions are asked of one-quarter of the monthly sample, and the annual release pools a year of those interviews. Only employed wage and salary workers are asked; the self-employed are outside the series. BLS notes that each January’s data reflect updated population controls, which can shift levels slightly from the prior year.

Members and represented

Two concepts are published. Union members are wage and salary workers who answer yes to being a member of a labor union or an employee association similar to a union. Workers represented by a union are the members plus those who answer no to membership but yes to their job being covered by a union or employee association contract. The union membership rate is members as a percentage of all wage and salary workers. State rates are published as well, with a caution that they rest on smaller samples and should be read with that in mind.

The earnings medians

The earnings figures are median usual weekly earnings of full-time wage and salary workers, defined as in the quarterly earnings release: pay before taxes and deductions, including overtime, commissions and tips usually received, at the main job, for people who usually work 35 hours or more. Medians are estimated from $50-wide intervals. The release gives one median for members, one for nonmembers, and a breakdown by characteristics such as sex, age, race, occupation group and industry.

What the gap reflects

The release states the caveat in its own text: in addition to coverage by a collective bargaining agreement, the difference between the two medians reflects differences in how union members and nonunion workers are distributed by occupation, industry, age, firm size and geographic region. Union membership is concentrated in particular industries and occupations, in the public sector, in larger establishments, in certain states, and among older workers, and each of those features is associated with pay on its own. A median over all union members is therefore a median over a differently composed workforce, and the raw gap mixes the effect of bargaining with the effect of composition. The release does not estimate how much of the gap each accounts for, and neither does this site.

What the tables cannot show

The series has no detail by occupation and area together, so it cannot say what unionized electricians earn in one state against nonunion ones. It has no measure of benefits, which collective agreements often address, and it reports earnings at the main job only. It is a snapshot of people who are members at the time of the interview, not of what happens to a given worker’s pay on joining or leaving. Changes from one year to the next are subject to sampling error that the release reports, and small movements are often not statistically significant.

Union status in OEWS

The occupation and state tables on this site come from OEWS, which asks employers nothing about union status. An occupation’s percentiles blend union and nonunion workers, and in occupations with substantial coverage the shape of the distribution partly reflects negotiated scales. The prevailing wage section is the one place where collectively bargained rates enter the site directly, since a Davis-Bacon determination is updated when a collective bargaining agreement changes the rate it reflects.

Reading the series well

The union release is a reliable account of how many workers are members or represented, where, and how their weekly earnings compare in the aggregate. The median weekly earnings series it draws on is explained in usual weekly earnings versus OEWS. What the release is not is an answer to “how much more does a union job pay,” a question that needs like-for-like comparison the published tables are not built to provide.