Pay rises faster where unions are present, and the federal numbers put a size on it. Full-time union workers earned a median of $1,294 a week in 2024 against $1,138 for nonunion workers, a difference of $156 (BLS, 2025). The gap explains why wage data, not just membership counts, is where labor economics gets argued.
This is a reading of the Bureau of Labor Statistics union membership file for 2024, released in January 2025, alongside the agency's compensation cost series. The point is not cheerleading for any side. It is that the same annual table carries both the premium and its limits, and the limits matter as much as the premium.
How large is the union wage premium?
The raw 2024 median difference is $156 a week, or roughly $8,100 a year for a full-time worker (BLS, 2025). The Bureau reports this as a simple comparison of medians, not a causal estimate. Union members skew toward older workers, longer job tenure, and mid-skill occupations in manufacturing, transportation and public service, all of which pay more regardless of a contract. Economists who adjust for those differences generally find a premium that shrinks but does not disappear, commonly estimated in the range of 10 to 20 percent.
The raw gap also moves with the composition of the unionized workforce. When union-heavy sectors shed jobs, the remaining members are more selectively employed, which can inflate the comparison without anyone getting a raise.
Where do unionized workers hold the biggest edge?
The occupations tell a sharper story than the national average. In 2024, the Bureau's data showed union concentration and the largest weekly-earnings differences in protective service occupations, education and training occupations, and production occupations in manufacturing (BLS, 2025). Public-sector coverage remains the anchor of the whole system: 32.2 percent of public employees were union members in 2024, against 5.9 percent in the private sector (BLS, 2025).
The two figures explain most of modern American labor policy. Any fight over public-employee bargaining law is a fight over a third of the unionized workforce. Any employer-side argument about union costs starts from the fact that less than six in a hundred private workers hold a contract.
What does the long membership decline do to the numbers?
The membership rate was 9.9 percent in 2024, covering about 14.3 million workers, down from 10.0 percent in 2023 (BLS, 2025). In 1983, the first year of the comparable series, the rate was 20.1 percent (BLS). The decline is not evenly distributed: it is concentrated in private manufacturing, where employment fell from about 19.6 million in 1979 to roughly 12.8 million in recent years (BLS payroll employment series).
The analytical consequence is understated. A shrinking, more selective union sector makes the raw premium harder to interpret, because the comparison group changes at the same time as the covered group. Statisticians call this a composition effect; it means the $156 weekly gap should be quoted with its caveat attached, every time.
Do contracts change pay growth, not just levels?
Level and growth are different questions. The Bureau's Employment Cost Index, which holds the mix of jobs constant, showed civilian compensation costs up 3.8 percent for the twelve months ending December 2024 (BLS, 2025). The ECI does not split union and nonunion growth in a way that supports a clean claim, and this article will not invent one.
What the evidence does support is narrower. Union contracts fix wages over multi-year terms, which flattens both spikes and slumps. That is why union pay tends to lag at the top of an inflation cycle and hold up better in a downturn. The 2024 medians capture a point in a cycle, not a trend line.
How should the premium be quoted responsibly?
Three habits separate a defensible use of the table from a misleading one. Quote the medians with the year, since the comparison moves annually. Name the measure — median usual weekly earnings of full-time wage and salary workers — because a household income figure or an hourly average is a different statistic (BLS, 2025). And keep the composition caveat attached, because the covered and uncovered workforces differ in age, tenure and industry in ways the raw medians never adjust away.
The same discipline applies to the sector rates. The 32.2 percent public and 5.9 percent private figures describe membership, not bargaining coverage; some workers are covered by contracts without being members, particularly in states that allow nonmember fee-payers. The two measures diverge by a few points, and analyses that swap them quietly produce different stories.
One concrete detail from the 2024 file is worth carrying: the gap is not uniform across the earnings distribution. Medians describe the middle, and the annual table's occupation detail shows union concentration in exactly the mid-wage occupations — production, transportation, protective service — where contracts historically set the local going rate rather than chase elite pay (BLS, 2025). That placement, more than any headline gap, is what the decline in private membership is slowly eroding.
What the 2024 numbers establish, and what they do not
The evidence establishes three things. Union members earned more per week in 2024, on median, than nonunion workers (BLS, 2025). Public-sector coverage remains several times private-sector coverage. And the covered workforce keeps shrinking slowly enough that the premium stays legible.
What remains unknown is the causal premium for any given worker, which depends on occupation, region and tenure in ways the annual table cannot resolve. Readers should treat the $156 figure as a description of the workforce as it exists, not a promise about what a campaign would produce. This publication reports the power, law and money around work without taking sides in a dispute; the table is quoted because it is the table.
Related: Algorithmic management at work · Reading employer health data · more in the workplace.
For more context, read Reading employer health data.
For more context, read Automation on the warehouse floor.
For more context, read Shift work and the body clock.