US employment data are hugely important. They cause large financial market moves and the Federal Reserve watches them closely because full employment is one of its goals.
But jobs numbers are also confusing. There isn’t just one measure of employment and different surveys often give different answers. On top of that, the numbers jump around from month to month and are regularly revised, sometimes a lot, long after they are first published. This means the headline number everyone reacts to is often not the final nor the most accurate one.
This article explains recent research that reduces confusion. The aim is to get a clearer picture of what is really happening in the US labour market, even when the official data are noisy, revised and delayed.
Our research approach* does three things. First, it keeps track of all past versions of the jobs data, so we can see how the numbers change over time. Second, it separates long-term forces, like population growth and migration, from shorter-term ups and downs. Third, it relies upon both private and public data so still produces useful estimates when official data are missing. Together, this gives a more stable and reliable view of the jobs market.
Breakevens: How many jobs does the economy really need?
A key question is how many jobs the US needs to add each month just to stop the unemployment rate from rising. This is sometimes called the ‘breakeven’ level of job growth. This breakeven changes over time. When the population is growing quickly or more people are entering the workforce, the economy needs to create more jobs. When population growth slows, fewer new jobs are needed to keep the unemployment rate steady.
Our estimates show that this breakeven level is falling fast. There is uncertainty around the exact numbers, but the direction is clear.
At the start of 2025, the economy needed to add around 150,000 jobs per month to keep the unemployment rate from increasing. Recently, that number has been closer to 70,000 and, by the end of 2026, it will fall further to 45,000 jobs per month.
The main reason is that population growth and migration into the US surged in 2023 and 2024 but are now returning to more normal levels. As a result, the labour force is growing far more slowly, and fewer new jobs are needed each month.
The bar for ‘strong enough’ job growth is much lower than it was just a year or two ago. This helps explain why the unemployment rate has started to edge higher even though job growth still appears positive.
Benchmarking: Why does employment get so heavily revised?
Another source of confusion is annual revisions to the employment data. The monthly payroll numbers are based on surveys, which are quick but imperfect. Once a year, they are adjusted using more complete records drawn from unemployment insurance data, which cover almost all workers but take much longer to collect. This process is called benchmarking.
This year’s preliminary revision suggests that employment levels will eventually be revised down. At first glance, the implied cut looks large. Simple methods that try to estimate the final revision early on are consistent with such a big revision.
Our analysis suggests a more moderate outcome. Based on past revision patterns, and additional data, we estimate that only about half of the initial gap is likely to remain in the final numbers. In other words, while some downward revision is likely, the worst-case interpretations probably overstate the true adjustment.
The key point is that early estimates of revisions for this year are too pessimistic and should be treated with caution.
Blindspots: What happens when data go missing?
Sometimes official jobs data are delayed or unavailable. Federal government shutdowns, reporting problems or other disruptions can leave markets without key information for weeks or months. Our framework is designed to handle these gaps. Because it continuously tracks the underlying jobs trend, it can still produce estimates even when official data are missing. In those cases, it relies more on private sources, such as automated payroll data and weekly unemployment claims reported by individual US states.
During the recent Federal government shutdown, when the official payroll report was delayed, our model continued to generate estimates. These suggested that monthly job growth was slowing toward around 40,000 jobs at the start of 2026. As this is below the breakeven level discussed earlier, unemployment was projected to gently rise toward about 4.5%. Both estimates have proved robust to the subsequent releases.
Conclusion
Job creation in the US is weakening, but the headline numbers can be misleading. Sharply falling population trends mean fewer new jobs are needed to keep the unemployment rate steady. Even so, job growth now appears to be running below that level, pointing to a gradual rise in unemployment into the start of 2026. Annual revisions are likely to lower reported employment, but probably by less than some early estimates have suggested. And when official data are delayed or revised, alternative indicators can still provide a useful guide.
For investors and policymakers, the main lesson is simple: don't take any single jobs report at face value. The real story only becomes clear when the noise is removed.
Jeremy Chiu and Daniel Wales are senior economists at Fulcrum Asset Management. The views expressed above should not be taken as investment advice.
* Full research paper: US Employment: Breakeven, Benchmarking and Blindspots by Chiu and Wales. Fulcrum Client Note, December 2025.