Manufacturing Payrolls

Manufacturing payrolls count factory jobs in Table B-1 of the Employment Situation, the monthly United States jobs report. The Bureau of Labor Statistics (BLS) builds the series from its monthly survey of businesses and agencies.

That sample covers about 1/3 of all nonfarm payroll jobs, with about 40 percent in establishments of fewer than 20 employees. It counts jobs paid for any part of the pay period including the 12th, so 1 worker with 2 jobs counts twice. Results come in thousands after seasonal adjustment for normal seasonal swings.

Collection began in October 1915 with four manufacturing industries, now coded under the 2022 NAICS. It counts jobs, not people or output, and leaves out the unincorporated self-employed, unpaid family and volunteer workers, farm and domestic workers, and military personnel. It says nothing about factory output, shipments, purchasing-manager indexes, unemployment, or headcount.

The reference is the pay period including the 12th, published about 3 weeks later, usually the first Friday after the reference month. The release stays under embargo until 8:30 a.m. Eastern Time.

It arrives with household-survey results, Tables B-1, B-2, B-3, B-7, and B-8, plus summary tables and highlights. First prints are preliminary, revised twice as more reports arrive. Each March re-anchors to tax-universe counts, rewriting 21 months of unadjusted data and 5 years of adjusted data.

Read the adjusted over-the-month change in thousands: about 130,000 counts as significant for total payrolls, versus about 600,000 in the household survey. No buy or sell level exists; 50 marks an even split between rising and falling industries.

Factory jobs move with vehicle swings, staffing depth, and hours rather than output itself.

• Durable goods: jobs in long-lasting goods industries like motor vehicles and parts and transportation equipment, the most volatile piece.
• Nondurable goods: jobs in short-lived goods industries, usually less volatile.
• Production and nonsupervisory employees: factory workers in fabricating, processing, assembling, and related work plus working supervisors.
• Average weekly hours: mean hours across all factory employees, read with jobs to gauge labor strain.
• Average weekly overtime hours: extra-hours gauge collected only from manufacturing plants, often shifting before jobs do.

The October 2023 round, released November 3, 2023, added 150,000 jobs, below the 258,000 prior-12-month average, while factories lost 35,000 and unemployment stood at 3.9%. Durable goods lost 36,000 jobs, including 33,200 in motor vehicles and parts on strike action, while nondurable goods added 1,000 and health care added 58,000. The factory workweek held at 40.0 hours, overtime slipped 0.1 to 2.9, and the factory diffusion index read 42.4.

August and September gains were cut 101,000 combined. The drop was strike-related, not a slump: workers unpaid the whole reference period do not count until they return.

1. Cell jobs estimate: Scale the prior month's jobs by the weighted change in firms reporting both months, excluding strike-hit reports, then add the birth-death forecast. For example, 12,995,000 times 0.997 gives 12,956,015, plus 3,000 gives 12,959,015, about 12,959 thousand, down about 36 thousand.
2. Birth-death adjustment: Impute new firms from closed ones inside the sample, then add a model forecast of the leftover gap from tax records.
3. Aggregation: Sum industry cells to the factory total and on to total payrolls, with worker-group ratios leaning about 90% on the prior month.
4. Hours and pay: Divide total hours by workers for weekly hours, and payroll by hours for hourly pay; overtime comes only from factories. Weekly pay is hours times pay. For example, 518,360,600 divided by 12,959,015 gives about 40.0 hours.
5. Seasonal adjustment: Apply X-13ARIMA-SEATS, Census software with fresh factors each month for the reference month and prior two. Durable and nondurable goods adjust directly, factory-wide overtime by aggregation, with calendar fixes for pay-period length and floating holidays.
6. Benchmark: Replace March with near-full tax counts and spread the gap over the prior 11 months. That rewrites 21 months of unadjusted data and 5 years of adjusted data.