The Bureau of Labor Statistic’s (BLS) Monthly Employment Situation Report, also known simply as the monthly jobs report, is one of the most highly-anticipated and closely-watched employment reports due to the timeliness of the data – released mere weeks after the survey data are collected. However, this timeliness entails a trade-off with accuracy, and that trade-off has been costlier in recent years, with both the monthly revisions (as more survey responses flow in) and the annual revisions (the result of benchmarking to other data sources) raising eyebrows lately.
For example, the 2024 annual benchmark revision estimates from August 2024 implied a larger than typical downward revision of over 800,000 jobs In August 2025, the seasonally adjusted employment estimates for June 2025 were revised downward by 27,000, flipping the estimate of job creation from positive to negative. The most recent benchmark revision estimates from September 2025 implied another larger than typical downward revision of over 900,000 jobs (-0.6 percent).
In this article, we break down how the estimates in the monthly jobs report are compiled, why the estimates undergo revisions, why those revisions have been so large recently, and whether the BLS data are trustworthy.
The Monthly Employment Situation – Two Surveys in One Report
The BLS’ monthly jobs reports combine data from two surveys: the Establishment Survey and the Household Survey. Both surveys produce sample-based estimates of employment, each with its own set of strengths and limitations.
The Establishment Survey
The BLS’ Current Employment Statistics program, also known as the Payroll Survey or the Establishment Survey, is a federal-state cooperative program that surveys approximately 121,000 non-farm businesses and government agencies each month, providing estimates of non-farm employment, hours, and earnings of workers on payrolls.
The Household Survey
The Current Population Survey, also known as the Household Survey, is a monthly survey of households conducted by the Census Bureau on behalf of the BLS. Based on responses to survey questions about work and job search activities, each person aged 16 and over in the sample is classified as employed, unemployed, or not in the labor force.
The Household Survey has a smaller sample size (roughly 60,000 households) but a more expansive scope than the Establishment Survey; it includes unpaid family workers, agricultural workers, self-employed workers whose businesses are unincorporated, and private household workers.
In addition to the differences in scope, the employment data collected by the Household Survey are also definitionally distinct from those collected by the Establishment Survey. The Household Survey counts employed persons, which can differ from jobs (as measured by the Establishment Survey) due to multiple job holders. In other words, two jobs as reported in the Establishment Survey may represent a single employed person in the Household Survey.
Revisions are by design, but are getting larger.
The Employment Situation Report is released early and often to help with decision-making. This entails a trade-off – because the jobs estimates in the report are survey-based, they’re not their best in their first release because those values are based on a sub-set of the full sample. These figures are preliminary and undergo revisions. Revisions are always going to happen, but they’ve gotten larger recently. We’ll look at three reasons for this: falling survey response rates, changes in immigration rates, and growth in gig work.
Falling Response Rates
Because the monthly jobs data are survey-based, the number of completed surveys received before each release deadline plays a large role in the size of revisions. The estimates from the Establishment Survey are revised twice in subsequent months as more survey responses flow in. The fewer responses there are to a survey:Non-response bias exists in every survey and the BLS employs strategies to both limit it and correct for it – but that becomes harder to do as response rates fall, as has been occurring across the globe and across all survey types. Various hypotheses have been put forward, including survey fatigue and mistrust of surveyors and/or government.
Source: Bureau of Labor Statistics
In terms of the Establishment Survey, the largest drop-off in response rates has been occurring in the first release of the employment estimates; it remains the case that most survey recipients have responded by the time of the final release.
Shifting Immigration Rates
The BLS aims to measure economic activity, not the legal status or workers. Therefore, the Establishment and Household Surveys do not ask about the legal status of workers and thereby are able to capture at least some employment of undocumented immigrants. The problem is that the QCEW data, against which the Establishment Survey data are annually benchmarked, excludes undocumented workers and business owners. While the BLS accounts for this exclusion from the QCEW data, it does so using an assumption of slow and steady immigration. Thus, during periods of large shifts in immigration rates, the benchmarking process may “over-correct,” leading to eye-catching revisions.
Rising Gig Work
Gig work (loosely defined as work mediated by online platforms rather than by employers) has emerged as a widespread phenomenon over the last decade, with the COVID-19 pandemic acting as an accelerant
Source: Internal Revenue Service
As self-employed independent contractors, platform workers are not subject to tax withholding and thus are not captured in the BLS’ QCEW data. Thus, while the BLS attempts to capture the gig workforce through the Contingent Worker Supplement (CWS) to the Household Survey, gig workers are excluded from the QCEW data, thereby making the current process of benchmarking to QCEW less effective and can lead to larger revisions – in much the same way as the surge in undocumented business owners and workers has.
Annual Benchmarking
Beyond the two monthly revisions based on additional survey responses that flow in, the employment estimates from the Establishment Survey are benchmarked annually to comprehensive counts of employment for the month of March from QCEW. These counts are derived from state unemployment insurance tax records that nearly all employers are required to file.
While the QCEW data are more rooted in hard data (administrative records from the IRS), QCEW does not cover some important groups like undocumented workers and independent contractors, both of which have undergone large shifts in recent years, as discussed above.
Trustworthiness of BLS Data
It is important to emphasize that the CES and QCEW measure different kinds of workers to varying degrees, so caution and in-depth knowledge of these differences must be used when making comparisons. Those differences notwithstanding, it cannot be denied that the BLS’ process of annually benchmarking CES estimates to QCEW data doesn’t appear to be working as well as it historically has, and may need to be adjusted to account for shifts in the labor force stemming from a global pandemic, shifting policies on immigration and trade, among other factors. First-release CES estimates could also be improved if more responses were received before the release deadline, but this would require more funding, which the Trump administration has ruled out.
Understanding the nuances and complexity of these reports and how they interact with one another brings to light areas of the process that need to be adjusted while dispelling any fear about manipulation of the data. As we’ve seen, revisions are inevitable, even under a better process; it’s simply not possible to get highly accurate data quickly in an economy as large as the U.S.’, with over 160 million workers. The agency publishes all changes to its methodology and is very transparent, allowing outsiders to check the quality of its work. For now, most economists say they still have faith in the BLS to try to put out the best, most accurate data it can.
IMPLAN’s Use of BLS Data
IMPLAN does not currently use any of the monthly jobs data. IMPLAN does rely heavily on the annual version of the QCEW data – but because the BLS’ QCEW data are based on administrative records from the IRS, rather than being model-estimated extrapolations of survey data like the monthly jobs estimates, the QCEW data undergo much smaller revisions, especially in regard to the annual values.
Because the QCEW contain various exclusions, IMPLAN augments the QCEW data with data from the Bureau of Economic Accounts, the National Oceanic and Atmospheric Administration, the Railroad Retirement Board, the Department of Defense, the Department of Agriculture, and more. Because transparency builds trust, we make our data sources and methodologies publicly available.