What do strong payroll numbers really tell us about the labor market?

The payroll employment growth estimate that BLS releases on Thursday, July 2^nd is the headline labor market number that markets, policymakers, and the media typically pay the most attention. There are good reasons for that. The payroll employment survey has a huge sample and is correspondingly precise about small percentage changes in total employment in any given month.1 

An underappreciated problem with interpreting these estimates, however, is that it can be hard to know what constitutes strong or weak growth. The three-month average of employment growth as of May 2026 was 188,000, as shown in Figure 1. Is that a lot or a little, in the sense of telling us that the labor market is strengthening or weakening? 

To think about answering that question, consider monthly payroll employment growth and the unemployment rate in recent years. Very high employment growth in 2022 was enough to continue pushing unemployment down, from 4.0% in January 2022 to 3.5% in January 2023. But growth remained high by historical standards throughout 2023—for example, higher on average than it was during 2019—while unemployment actually ticked up to 3.7% by January 2024. Was employment growth strong or weak in 2023? 

There is some uncertainty about any estimates of breakeven, and consequently there are a variety of estimates at any given time. Figure 3 shows some of those estimates for 2026, including the estimate from Figure 2. As indicated by the high and low immigration scenarios considered by Edelberg, Veuger, and Watson (2026), some of the variation is due to uncertainty about immigration flows. Another source of uncertainty is the unknown lag times between changes in immigration flows and changes in employment: many immigrants are not immediately eligible for work or simply need time to enter the labor force. Yet another issue is the shifting availability of residents for work. Population aging has predictable consequences for that availability, but (for example) the introduction of remote work may have boosted the participation rates of individuals with disabilities and mothers of young children. This would tend to raise the breakeven rate without any change in the population size.

UI claims can provide a very timely signal of labor market distress, especially when it is accompanied by a large spike in layoffs. But there are downsides to relying on UI measures. An important downside is rooted in the fact that UI benefits are only received by a small minority of the unemployed (as defined in worker survey data), with wide variation in that recipiency rate across states and time. Unemployment insurance data is worth tracking as part of an overall labor market assessment, but its limitations should be kept in mind.

Another labor market metric is the monthly unemployment rate. It is estimated from nationally representative worker survey data rather than UI records and is just slightly less timely than UI data. (On Thursday July 2^nd, we will learn the unemployment rate for mid-June.) Figure 5 shows the national unemployment rate (again omitting the pandemic period) over the last ten years. Like continued claims, the unemployment rate fell during the late 2010s, going from its post-Great Recession height of 10.0% in October 2009 all the way to 3.6% in December 2019. It began to rise somewhat later than claims, from a trough of 3.4% in April 2023 to its current level of 4.3% as of May 2026.

Importantly, the unemployment rate covers the entire population. It has been measured in a consistent way for decades, identifying the nonemployed workers with the clearest indications of availability and desire for work. Historically speaking, spikes in the unemployment rate have correlated closely with recessions. And by contrast to payroll employment growth, the unemployment rate is not as affected by shifts in immigration and other factors that confuse its interpretation.

Like any measure, the unemployment rate has disadvantages as a labor indicator. One is that it does not capture everyone who is potentially available for work; every month there are many labor force nonparticipants who directly transit to employment without ever showing up in the survey data as unemployed. The household survey that generates the unemployment rate is also based on a smaller sample than the payroll survey. That sample is constructed to be nationally representative, but declining response rates and nonresponse bias specifically can affect its interpretation. For example, to the extent that over time undocumented immigrants become more or less likely to answer the survey, this could marginally affect the measured unemployment rate.