Much knowledge about labor markets in the United States comes from employer reports, such as employers' responses to government surveys and tax filings. While employer responses provide valuable labor market information, there are limitations on what employers can or will report about changes in labor markets.
This In Focus describes measures of labor markets based on employer reports, with an emphasis on the value and limitations on information from employers about how Artificial Intelligence (AI) and Generative AI (GenAI) technologies are now and may in the future affect labor markets. It relies on ideas presented in CRS In Focus IF13182, The Impact of Generative AI on Labor Markets: Frameworks.
Many existing measures of labor markets—including measures critical for understanding the impacts of AI and GenAI—are based on data collected from employers.
Some of the most fundamental concerns about how AI and GenAI may affect labor markets involve potential changes in employment levels. Changes in the number of jobs in the United States are measured each month by the Bureau of Labor Statistics (BLS) Current Employment Statistics (CES) program. The CES surveys thousands of employers each month, asking how many people are on these employers' payrolls. These employer reports are used to estimate employment changes (total and by industry) with less than a one-month lag. CES employment change estimates are benchmarked each year to incorporate less timely but more comprehensive data based on employers' unemployment insurance tax filings, as compiled in the Quarterly Census of Employment and Wages (QCEW). These data do not capture reasons for employment changes. Some private sector payroll companies also publish estimates of employment change based on payroll data.
New technologies, such as AI and GenAI, may lead to changes in employment for particular types of work, which are not captured in the CES or QCEW. The Occupational Employment and Wage Statistics (OEWS) program at BLS publishes estimates of employment by occupation for each state and metropolitan area annually. OEWS estimates are based on surveying hundreds of thousands of employers about their employment levels by occupation and combining these employer responses with QCEW information on the full population of employers. Each OEWS estimate is based on three years of data collection, so changes in employment by occupation can be estimated only with these data over three-year periods. These data also do not capture reasons for employment changes.
Overall employment levels change through hiring and separations (e.g., layoffs and retirements). The BLS Job Openings and Labor Turnover Survey (JOLTS) measures hires and separations by surveying employers about their hires and separations in the previous month and combining these employer survey responses with QCEW information on the full population of employers. JOLTS estimates are available by state every month. Some policymakers and observers concerned with measuring the impacts of AI and GenAI have suggested adding occupational questions to JOLTS to track changes in hiring and separation by job type more rapidly than new OEWS estimates are available.
Some researchers have used employer data from payroll providers and administrative tax records to study changes in hiring patterns for young workers by occupation or by industry and whether or not these changes can be explained by new technologies. Other researchers have examined employers' online job postings for junior roles to study changes in labor demand that may be due to these technologies. Some observers have suggested more states should collect occupational data from employers as part of their unemployment insurance tax filings to better measure changes in hiring and separation patterns.
The Census Bureau's Business Trends and Outlook Survey (BTOS) surveys businesses every two weeks. BTOS questions added in 2025 ask these businesses if they are using GenAI, how they use AI and GenAI, and whether their use of AI is increasing or decreasing their overall employment. Census Bureau surveys can be linked with longitudinal data on employment patterns for these businesses. Other surveys asking businesses about their current and expected future AI use have been sponsored by regional Federal Reserve banks.
New technologies can change what workers do on the job. People with the same job title in different time periods may do different tasks during their workdays.
BLS surveys employers to ask what people do on the job in specific occupations through the Occupational Requirements Survey (ORS). This survey, sponsored by the Social Security Administration to inform disability determinations, focuses on the physical and cognitive requirements of jobs. Some observers have suggested ORS could collect more information on how jobs are changing as AI and GenAI are deployed in workplaces.
Employer-provided data are valuable but have limitations to capturing the impact of new technologies on labor markets. Limitations of employer-provided data include the timeliness of published data and the contemporaneousness of classification systems. Other limitations involve the accuracy and representativeness of employer reports and limitations in the information employers can provide.
The CES program produces estimates of month-to-month changes in employment with less than a one-month lag, followed months later by more comprehensive QCEW employment estimates. Both CES and QCEW estimate changes in employment by industry—the classification of what businesses produce overall (e.g., manufacturing or health care)—and not by the types of work people do within these businesses (e.g., machine repairing).
New technologies may change employment levels for specific types of work. However, changes in employment by industry do not always measure changes in employment by type of work. For example, the contracting out (domestic outsourcing) of factory production jobs in the United States in the 1990s from manufacturing industry employers to employment services industry employers was large enough to affect whether measured manufacturing industry production jobs increased or decreased during this period.
Millions of U.S. workers work in occupations that did not exist 30-40 years ago. One challenge in measuring new types of work is that employment in new occupations is reported only by the federal statistical system as occupations are added to the Standard Occupational Classification (SOC) system. For example, "Data Scientists" were added in the 2018 SOC update. Before this update, Data Scientist employment was included in the employment of "Mathematical Occupations - All Other." In 2024, the SOC Policy Committee began the first SOC update since 2018, for a version of the SOC scheduled for use beginning in 2028. Some observers have suggested the SOC should be updated more frequently to measure the impact of new technologies on labor markets.
In 2025, employers cited AI in laying off thousands of workers. Outside observers suggested that these employers may have overstated the impacts of AI on layoffs to appear innovative rather than ascribing the layoffs to slowing sales, previous over-hiring, or negative impacts of policy changes. Overstating the impacts of AI on business decisions has been called "AI-washing."
Employers also may have reasons to understate the labor displacement impacts of AI. These reasons could include the reduction of employee opposition to new technology or concerns regarding proposals to tax AI adoption.
Employer response rates to federal surveys have been falling in recent years. (The BTOS began in 2022 and has lower response rates than other federal statistical programs mentioned above.) Low response rates mean estimates based on survey data can be affected by variability in which employers choose to respond to the surveys, in addition to the actual changes in labor markets that these surveys are intended to measure. In contrast, federal and state laws require nearly all employers to file the tax reports underlying the QCEW, which has higher reporting rates.
Many labor market measurement programs—including the CES, OEWS, JOLTS, ORS, and BTOS—collect data from employers. However, not all workers work for employers. Some studies of how GenAI is affecting freelance labor markets have used data from platforms for buying and selling illustrations and stock images.
New technologies can affect labor supply and labor demand. For example, the increase in home ownership of washing machines in the mid-20th century may have increased U.S. married women's availability for work outside their homes. There is some evidence now that people are increasingly using GenAI for nonwork tasks. If this new technology use affects labor supply, employers would not be the best source of information about it.
The full impact of technology on employers' labor demand is determined by their own technology adoption decisions and by technology adoption decisions made by their customers, suppliers, and competitors. Thus, asking employers how technology affects their own employment decisions cannot capture the full impact of how technology affects labor markets.
As an example of this measurement problem, in the early 2000s, BLS asked employers involved in mass layoffs (those resulting in at least 50 unemployment insurance claims) about the cause of each mass layoff. From 2004 through 2011, these employers reported laying off an average of 11,000 workers per year because of work being moved out of the United States. Later, more comprehensive analyses examined changes in global trade patterns by industry and the relationship between increased imports to the United States and changes in employment within the United States. These analyses showed increased imports from China were associated with reductions in about 76,000 U.S. manufacturing jobs per year during this period.
Asking employers to identify the causes of mass layoffs did not fully capture the labor market impacts of greater trade with China earlier this century and is unlikely to fully capture the impacts of AI and GenAI technologies today. Instead, capturing these impacts requires measures of (1) how new technology is being adopted, (2) how the adoption of new technology is changing what workers do on the job, and (3) changes in employment by type of work (occupation).