Household Survey vs Establishment Survey: Two Jobs Reports

The household survey (Current Population Survey) asks individuals whether they are employed; the establishment survey (Current Employment Statistics) asks businesses how many people they employ. Both measure labor supply, but diverge significantly because one counts workers and the other counts payroll jobs—and the reconciliation reveals economic churning, gig work, second jobs, and mismeasurement.

The Two Surveys: What They Count

The household survey (Current Population Survey, or CPS) interviews roughly 60,000 U.S. households each month. Surveyors ask: “Did you work for pay last week?” If yes, you are employed. The survey counts people, regardless of how many jobs they hold or how formal the arrangement. A freelancer, an Uber driver, a person with two part-time jobs—all count as one employed person.

The establishment survey (Current Employment Statistics, or CES) surveys roughly 400,000 business payroll records. It asks: “How many employees were on your payroll in the reference week?” It counts jobs, not people. A person with two part-time jobs appears as two jobs in the CES.

Both are released monthly by the Bureau of Labor Statistics. Both are noisy; both are revised. Yet the public fixates on a single “jobs number,” often the establishment survey payroll-job count, because it is cleaner and feeds into Fed policy.

This is a mistake. They measure different things, and the gap between them tells you something.

Why They Diverge

Over the past decade, the household survey has often counted more employed persons than the establishment survey’s job count would suggest. If the CES reports 200 million payroll jobs and the CPS reports 158 million employed persons, the gap implies that either (1) many people hold multiple payroll jobs, (2) many employed people are not on formal business payrolls (self-employed, gig, informal), or (3) one survey has a significant error.

The main culprits are:

Self-employment and gig work. The CPS includes all self-employed people (consultants, sole proprietors, contractors, rideshare drivers, freelancers). The CES excludes self-employed persons entirely—it only counts payroll employees. Over the past 15 years, gig work has expanded, and the CPS might be capturing growth in this segment while the CES misses it.

Multiple job holders. The CPS counts a person with two jobs as one employed person. The CES counts two jobs. If the share of workers with multiple jobs rises, the CES will grow faster than the CPS, ceteris paribus. Conversely, if workers shed second jobs, the CES falls while the CPS might stay flat or fall less.

Business births and deaths. New, very small businesses and side ventures (especially sole proprietorships) may not appear in the CES sample immediately. The household survey catches them sooner.

Misclassification. Some people classified as self-employed (e.g., contractor invoicing a company) might argue they are employees. The IRS and labor courts have wrestled with this (classifying Uber drivers, for instance). Survey respondents themselves may be ambiguous about their status.

Sampling and revision error. The CES is revised annually based on comprehensive administrative payroll data (UI records). Large revisions can occur, revealing that the sample was off. The CPS is also revised, but usually by smaller amounts.

Historical Examples

In late 2022 and early 2023, the CES was reporting robust job growth (300,000+ per month), while the CPS household employment growth slowed. Some economists argued the CES was overstating labor market strength. When the annual CES revision came in mid-2023, it showed that payroll growth had been weaker than initially reported—a revision of several hundred thousand jobs. The gap had partially reflected mismeasurement, not just divergent trends.

Conversely, in the early COVID recovery (2020–2021), the CPS household employment recovered faster than the CES payroll counts suggested, raising questions about whether people were re-entering work via informal channels or whether the CES was slow to capture hiring.

Which Is More Reliable?

Neither is definitively “more reliable.” The answer depends on your question.

For measuring total labor force participation, the CPS is better. It captures the full spectrum of work, including self-employment and gig, and does not double-count workers with multiple jobs.

For measuring formal business payroll risk and hiring momentum, the CES is better. It is based on actual payroll records submitted by employers for tax purposes, not self-reported survey responses. A business cannot hide payroll jobs; it must report them to the IRS.

For early-cycle signals, the CES often leads. Businesses hire payroll employees before contractor demand picks up, so CES growth often precedes broader labor market expansion.

For accuracy on the unemployed and labor force entry, the CPS is essential. It defines unemployment, labor force participation, and tracks non-participants. The CES says nothing about people not working.

The Convergence Problem

Policy-makers often speak of “jobs created” as if there is one number. In fact, the relevant metric depends on context. When the Federal Reserve worries about labor market overheating and wage inflation, it cares about whether people are finding work and bidding up wages—a household survey question. When equity analysts worry about corporate payroll costs and earnings power, they care about payroll jobs—an establishment survey question.

Conflating the two can lead to policy errors. If the CES is strong but the CPS is weak, it might reflect growing gig work and part-time multiple jobs—which could be a sign of labor market fragmentation and declining job quality, not strength. Conversely, if the CPS is strong but the CES is weak, it could mean workers are leaving formal payroll employment for self-employment or informal work—a shift in type of work, not necessarily in total employment.

The most honest summary: compare them both, watch the gap, and ask why it exists.

See also