Employment Reports: Nonfarm Payrolls and Household Survey

Two Surveys, One Report (Why You Need Both)

The Bureau of Labor Statistics releases the Employment Situation report on the first Friday of each month at 8:30 AM Eastern. This single report moves Treasury yields, equity futures, and Fed rate expectations more consistently than any other economic release. In one study of high-frequency market reactions, nonfarm payrolls surprises moved 10-year Treasury yields by 3.5–5.0 basis points per 100,000-job surprise within minutes of release (Fleming & Remolona, 1999).

The report contains data from two separate surveys that measure different things and sometimes tell conflicting stories. If you only read the headline number, you're using half the data—and often the less informative half.

The point is: the employment report is not one number. It's a system of cross-checks, and the gaps between surveys frequently contain more signal than the headlines.

---

What Each Survey Actually Measures (The Definitions That Matter)

The establishment survey (nonfarm payrolls)

The BLS surveys approximately 670,000 worksites covering about one-third of all nonfarm payroll employees. This is the source of the headline "economy added X jobs" number. It counts payroll jobs, not people—if you hold two jobs, you appear twice.

What it captures:

• Total nonfarm employment (the headline number)
• Average hourly earnings (the wage inflation signal)
• Average weekly hours worked (a leading indicator of labor demand)
• Employment by industry sector

What it excludes:

• Agricultural workers
• Self-employed individuals (roughly 10 million workers)
• Unpaid family workers
• Private household employees

The household survey

The BLS surveys approximately 60,000 households and classifies each working-age individual as employed, unemployed, or not in the labor force. This is where the unemployment rate comes from. It counts people, not jobs—if you hold two jobs, you appear once.

What it captures:

• Unemployment rate
• Labor force participation rate
• Employment-to-population ratio
• Part-time vs. full-time employment breakdown
• Duration of unemployment
• Self-employment and agricultural employment

The point is: these surveys answer fundamentally different questions. The establishment survey tells you how many jobs exist. The household survey tells you how many people are working (and how many want to work but can't find jobs). Both answers matter, and they're not interchangeable.

---

How the Key Calculations Work (With Numbers)

Unemployment rate

The formula:

Unemployment Rate = (Unemployed / Labor Force) × 100

Where Labor Force = Employed + Unemployed (people actively seeking work in the past 4 weeks).

Example with real numbers:

• Civilian labor force: 168.0 million
• Employed: 161.5 million
• Unemployed: 6.5 million
• Unemployment rate: (6.5 / 168.0) × 100 = 3.87%, rounded to 3.9%

Why this matters: the denominator (labor force) moves too. If 500,000 people stop looking for work, the unemployment rate can fall even though no one found a job. The unemployment rate can improve for bad reasons. That's why you always check labor force participation alongside it.

Labor force participation rate

The formula:

Participation Rate = (Labor Force / Civilian Noninstitutional Population) × 100

The current participation rate (roughly 62.5–62.8%) is well below the pre-2008 peak of ~66.4%. Demographics explain part of this (Baby Boomers retiring), but not all of it. A rising participation rate alongside falling unemployment is the healthiest signal—it means the economy is pulling people off the sidelines.

Average hourly earnings

This comes from the establishment survey and is the primary wage inflation metric in the report. Year-over-year growth above 4.0% signals wage-driven inflation pressure that the Fed watches closely. Month-over-month readings above 0.4% (which annualize to roughly 5%) tend to move bond yields higher on release day.

---

Sector Composition (Where the Jobs Are)

Not all payroll gains are equal. The composition tells you more about economic quality than the headline number.

The practical point: a report showing +200,000 headline jobs with gains concentrated in government and healthcare tells a very different story than +200,000 driven by professional services and manufacturing. Read the sector detail before forming a view.

---

Worked Example: October 2024 (How Distortions Hide the Signal)

This example shows why reading beyond the headline is non-negotiable.

The headline: October 2024 reported +12,000 nonfarm payrolls, dramatically below the +100,000 consensus expectation. At first glance, this looked like a sharp labor market deterioration.

The context you needed:

• Hurricane Helene struck the Southeast in late September, with ongoing disruptions through October
• Hurricane Milton hit Florida in early October
• ~44,000 Boeing workers were on strike during the survey reference week
• The BLS noted that hurricanes and the strike reduced payrolls by approximately 100,000 combined

The adjustment:

• Reported: +12,000
• Estimated weather and strike impact: ~100,000
• Underlying trend: approximately +112,000 (close to the slowing-but-healthy trend)

What markets actually did: Initial futures sold off on the headline, then partially recovered as traders read the details. Treasury yields fell modestly, consistent with a "distorted, not disastrous" interpretation.

The lesson worth internalizing: any single month's payrolls can be distorted by weather, strikes, government shutdowns, or seasonal adjustment quirks. You never trade a single month's headline in isolation. The three-month moving average strips out most noise—and that's the number the Fed emphasizes in its communications.

---

Why the Two Surveys Diverge (And What It Signals)

In 2022–2023, the household survey showed notably weaker job growth than the establishment survey. This divergence persisted for months and triggered heated debate about which survey was "right."

Five reasons the surveys diverge:

1. Multiple jobholders: If gig work and second jobs increase, the establishment survey counts each job separately while the household survey counts each person once. Rising multiple-jobholding inflates the payroll count relative to the household count.

2. Self-employment trends: A shift toward self-employment (freelancing, contract work) shows up in the household survey but is invisible to the establishment survey. Post-pandemic, self-employment ran roughly 500,000–800,000 above pre-2020 trends.

3. Immigration and population estimates: The household survey relies on Census population controls that may lag actual immigration patterns. If population growth is underestimated, the household survey understates employment growth.

4. Birth-death model adjustments: The establishment survey uses a statistical model to estimate net job creation from new businesses minus closures. This model uses historical patterns and can be slow to capture turning points (more on this below).

5. Sample size differences: The establishment survey covers 670,000 worksites versus 60,000 households. The payroll data has a much smaller confidence interval around any given month's estimate.

The point is: neither survey is "wrong." They measure different things with different methods. Large, sustained divergences (lasting 6+ months) often signal structural shifts in how people work—more freelancing, more multiple jobholding, or population changes that statistics haven't caught up with. When you see a persistent gap, investigate which component is driving it rather than dismissing one survey.

---

Revisions and Benchmark Adjustments (The Numbers Change After Release)

The establishment survey undergoes significant revisions that can reshape the narrative months or even a year later.

Worked example: The preliminary March 2024 benchmark revision showed that total payroll gains for the 12 months ending March 2024 were revised down by 818,000 jobs—an average overstatement of roughly 68,000 per month. This was the largest downward benchmark revision since 2009 and significantly recast the 2023 labor market narrative from "resilient" to "cooling faster than initially reported."

Why this matters for investors: if you positioned based on the initially reported "strong" payroll numbers throughout 2023, the benchmark revision revealed the underlying reality was meaningfully weaker. Revisions are not footnotes—they're core data. Always check whether the prior two months were revised up or down when reading the current release.

---

The Birth-Death Model (Why Turning Points Are Tricky)

The BLS uses a "birth-death model" to estimate the net effect of new business formation minus business closures on total employment. This adjustment is necessary (the establishment survey can't survey businesses that don't exist yet), but it introduces systematic errors at economic inflection points.

How it works:

• The model uses historical seasonal patterns of business creation and closure
• During expansions, new business formation typically exceeds closures
• The adjustment adds (or subtracts) estimated jobs from the raw survey data

Where it breaks down:

• During recessions: the model continues adding jobs based on historical expansion patterns, even as business closures accelerate. This can overstate employment by 50,000–100,000 per month in the early stages of a downturn (before benchmark revisions correct it).
• During recoveries: the model may undercount new business formation, especially in emerging industries not well-represented in historical patterns.

The practical point: in the months surrounding economic turning points, treat the birth-death adjustment as a source of systematic error, not a precision tool. The annual benchmark revision (which uses actual unemployment insurance tax records covering nearly all employees) corrects these errors—but the correction comes 9–15 months after the fact.

---

What Moves Markets on Employment Friday

Components ranked by typical market impact:

1. Headline payrolls vs. consensus: The surprise direction and magnitude drive the initial reaction. A miss of ±50,000 or more from consensus typically moves 10-year yields by 5–10 basis points and equity futures by 0.3–0.8% within minutes.

2. Average hourly earnings (year-over-year): Growth above 4.0% signals wage inflation and tilts Fed expectations hawkish. Growth below 3.5% supports the disinflation narrative.

3. Prior month revisions: A headline of +180,000 with +60,000 in upward revisions to prior months tells a much stronger story than +180,000 with -40,000 in downward revisions. Always net the revisions into your read of the report.

4. Unemployment rate changes: Moves of 0.3 percentage points or more in either direction are statistically and economically significant. Moves of 0.1% are within sampling error (the 90% confidence interval on the unemployment rate is approximately ±0.2 percentage points).

5. Labor force participation rate: Rising participation alongside falling unemployment is the most bullish combination. Falling participation that flatters the unemployment rate is a yellow flag.

6. Hours worked: Payrolls can rise while total hours worked fall (shorter workweeks). Declining average weekly hours is an early signal of slowing labor demand—employers cut hours before cutting headcount.

---

Common Pitfalls (And How to Avoid Them)

Pitfall 1: Reacting to the headline without reading the details. A single month's payroll number has a 90% confidence interval of approximately ±130,000 jobs. A report of +150,000 is statistically indistinguishable from +20,000 or +280,000 in a single month. Use the three-month moving average for trend, not any individual print.

Pitfall 2: Ignoring seasonal adjustment distortions. January typically involves large seasonal layoffs (holiday workers leaving), and the seasonal adjustment can swing the headline by 200,000+ depending on whether the pattern matches historical norms. Summer months face similar issues with education-sector employment.

Pitfall 3: Treating small unemployment rate changes as meaningful. A move from 3.7% to 3.8% is within sampling error. Only moves of 0.2–0.3 percentage points or more (especially when sustained across 2–3 months) indicate a genuine shift.

Pitfall 4: Missing the hours-worked signal. If payrolls rise by +200,000 but average weekly hours fall from 34.4 to 34.1, total labor input actually declined. Hours worked × payrolls = total labor input, and this composite is a better real-time measure of economic momentum than either component alone.

Pitfall 5: Ignoring the household survey when it diverges. The household survey's employment-to-population ratio and part-time-for-economic-reasons data add context that payrolls alone cannot provide. When the two surveys tell different stories for 3+ months, the divergence itself is information.

---

Employment Report Checklist

Before 8:30 AM (preparation)

• Know consensus expectations for headline payrolls, unemployment rate, and average hourly earnings
• Check for known distortions: strikes, hurricanes, government shutdowns, Census hiring
• Note recent Fed commentary on labor market conditions (especially references to specific metrics)
• Review the prior month's revisions to calibrate expectations for this month's revision direction

After the release (first 10 minutes)

• Compare headline payrolls to consensus and note the surprise direction
• Check revisions to the prior two months (net them into your headline read)
• Calculate the three-month moving average for the trend signal
• Check average hourly earnings year-over-year and month-over-month
• Examine unemployment rate change and whether it was driven by job gains or participation shifts
• Review employment-to-population ratio from the household survey
• Check average weekly hours for early signals of labor demand changes

Trend assessment (within 24 hours)

• Compare establishment and household survey directions—do they agree?
• Review sector composition: which industries drove the gains or losses?
• Assess whether the report changes the Fed's likely path (check Fed funds futures after the release)
• Update your three-month and six-month trend estimates

---

Next Step

For the next three employment reports, track both the establishment and household survey employment changes side by side. Calculate the three-month moving average of payrolls after each release and note the revision direction. When the two surveys diverge significantly, investigate which components (self-employment, multiple jobholders, participation shifts) are driving the difference. Building this habit turns the employment report from a headline reaction into a systematic macro signal.