Two facts about American unions get published every year, usually within a week of each other, and they are used to argue opposite things.
The first is that organising is having its best run in a generation. Workers winning representation elections at the National Labor Relations Board rose 19 percent last year to 93,517, the highest count in decades.
The second is that union density is 10.0 percent — 14.7 million members — having moved up a tenth of a point from 9.9, in a series that has risen year on year only three times since 1980.
Neither figure is wrong and neither is being spun. They differ because one counts a flow and the other measures a stock, and the relationship between them is arithmetic that nobody quoting either number tends to do out loud.
Do the division
If 14.7 million members are 10.0 percent of covered workers, the workforce is about 147 million.
One percentage point of density is therefore about 1.47 million members.
A record organising year delivered 93,517 election winners. Against 1.47 million, that is about six hundredths of a percentage point.
And 93,517 is the optimistic end. Winning an election is not joining a union in the sense the density statistic counts; a first contract still has to be reached, and a substantial share of newly certified units never get one. The number that reaches the membership figure is smaller than the number that makes the headline.
The denominator is also moving
Standing still is not free.
If total employment grows and union membership does not grow proportionally, density falls without a single member being lost. At 10 percent, holding the line requires organising roughly one in every ten net new jobs — which, in an economy adding a million or more jobs in a good year, means about 100,000 new members annually before any attrition is replaced.
Attrition is not small. Union members retire. Unionised workplaces shed people, and this desk's past three days are a catalogue of it: 4,000 posts going at Jaguar Land Rover, and a European trade body putting 300,000 manufacturing jobs at risk before December. Employment growth, meanwhile, is concentrated in sectors where density is lowest, which means the jobs being added are structurally harder to organise than the jobs being lost.
So a record year of roughly 93,500 election wins, of which some fraction converts to membership, is running against a treadmill calibrated at around 100,000 a year just to keep the rate flat. That is why a 19 percent increase in organising produced a one-tenth-of-a-point move in density, and it is why the tenth of a point is itself unusual.
Where the ten percent actually is
The aggregate hides a split so wide that the average describes almost nobody.
Public sector: 32.9 percent. Private sector: 5.9 percent.
Those are different countries. A third of public employees are represented; one private worker in seventeen is. The highest private-adjacent rates — utilities at 17.8 percent, transportation and warehousing at 13.6 — are in industries with fixed physical assets that cannot be relocated to avoid a bargaining unit. Education, at 13.4 percent, is mostly public.
The pattern is not about worker sentiment, which surveys have shown running well ahead of membership for years. It tracks whether the employer can move. That is a structural fact about capital, not a fact about organising skill, and no amount of the second changes the first.
What the numbers do not settle
This is deliberately an argument about arithmetic and not about whether unions are good, which is a question the division cannot answer.
What the arithmetic does establish is that the current organising model — unit by unit, election by election, employer by employer — cannot move national density, and that this would remain true if it ran twice as well as its best year in decades. Anyone claiming a revival on the strength of the election numbers is describing something real that is nonetheless too small to show up. Anyone dismissing the movement on the strength of the density number is describing a denominator, not an effort.
Both sides are looking at the same treadmill and reporting the part they can see.
What to watch
Not next year's density print, which will land within a tenth of a point of this one whatever happens, because that is what a stock of 14.7 million does.
Watch first-contract rates at newly certified units — the share that reach an agreement within two years. That is the conversion ratio between the flow everybody celebrates and the stock everybody measures, and it is the only number in this system that could change by enough to matter without anybody organising a single additional worker.
The union membership rate of 10.0 percent in 2025 against 9.9 percent in 2024; the figure of 14.7 million union members among wage and salary workers; public-sector density of 32.9 percent and private-sector density of 5.9 percent; the sector rates for utilities, transportation and warehousing and educational services; and the observation that density has risen year on year only three times since 1980 are as published by the US Bureau of Labor Statistics in its union members release of February 2026 and as compiled from it. The figure of 93,517 workers winning NLRB representation elections, up 19 percent from 78,791 and described as the highest in decades, is as compiled from National Labor Relations Board election data. The implied total of about 147 million wage and salary workers is derived from the BLS membership count and rate. The arithmetic relating election wins to density, and the analysis, are our own.
Topics nationallabourworklabor day
