Recent payroll data for July from the Bureau of Labor Statistics (BLS) shows some weakness in the labor market, especially after large downward revisions for prior months. President Trump reacted strongly to this release, firing the BLS Commissioner and claiming that this and past data have been rigged against Republicans. One possible check on the payroll statistics comes from another BLS data base—the household survey used to calculate the unemployment rate. Although the two surveys have different statistical properties and methods, somewhat different definitions of employment, and entirely different data sources, they are highly correlated and, over even relatively short periods, show largely similar results. What does the household survey now show compared to the payroll establishment survey?
Both surveys produce sample-based estimates of employment. The establishment survey of payrolls of nonfarm businesses and government agencies has a much larger sample size and therefore a smaller margin of error for measuring month-to-month changes. Moreover, as additional sample receipts come in from survey respondents and seasonal adjustment factors are recalculated, the prior two months’ data are revised to improve accuracy. There are no monthly revisions to the data from the household survey, and its smaller sample size results in a much larger margin of error. The household survey has a more expansive scope, including self-employed workers whose businesses are unincorporated, unpaid family workers, agricultural workers, and private household workers, who currently represent more than two percent of total employment. In the household survey, workers in the civilian non-institutional population above age 16 are only counted once, regardless of the number of jobs they hold. In contrast, the establishment survey counts the total number of nonfarm, nonmilitary jobs held by people of all ages. The household survey is benchmarked annually to the Census estimate of the population, while the establishment survey is benchmarked annually to premium payment counts from state unemployment insurance programs. Note that because of an unusually large increase in population estimates in the official January 2025 household survey statistics that actually represents a longer-term increase arising from net immigration and methodological changes, I use below an alternative BLS data series from April 2020 through December 2024 for the household survey that better accounts for the increase.
Chart 1 shows the seasonally adjusted monthly changes in employment according to the establishment and household surveys since January 2017, with the wide swings of the pandemic months from March 2020 to October 2020 placed in a separate box. As can be seen, the two series generally track each other closely (the contemporaneous correlation is 0.980), though the household survey is more volatile. Its standard deviation, excluding the pandemic recession months, is 814 thousand compared to 565 thousand for the establishment survey. Summing the monthly changes over a calendar year, and more so over four-year presidential administrations, yields largely similar results for the two series. For example, in 2018 Table 1 shows the household survey recorded 2.9 million additional workers, while the establishment survey recorded 2.3 million new jobs. Over the first Trump Administration, which includes the deep but brief 2020 pandemic recession, the comparable figures are -2.1 million and -2.7 million.
Chart 2 shows the monthly changes in employment for the two surveys over the period from February 2025 to July 2025. Although the household survey is more volatile, both indicate weak or even negative employment growth. Over the entire period, the household survey records 0.8 million fewer workers, while the establishment survey shows 0.5 million new jobs, of which only 100 thousand were added in the last three months, mainly in the health care sector.
In short, two different surveys show similar weak results for the recent employment situation. These data are important to federal policymakers in formulating monetary and fiscal policies, as well as to businesses and households evaluating the economic environment in which to invest, work, and spend. It is therefore vital that the statistics be based on the best data and methodologies available, without any taint of political interference. That being said, voluntary initial responses to surveys have declined in recent years, and government budgets are tight. The BLS, assisted by other federal and state agencies such as the IRS, SSA, and the Department of Labor, should reconsider all concepts and sources used in formulating its employment statistics. This should include drawing upon government administrative and private sector data in addition to surveys and speeding up and tightening required reporting, for example, from unemployment insurance programs and payroll tax collections. These changes could also benefit the administration of government social insurance and welfare programs, to improve benefit accuracy, by reducing over- and under-payments.
The post Recent Different Source Statistics on the Current Employment Situation Confirm Weakness appeared first on American Enterprise Institute – AEI.
