Take the total weekly payroll paid to America's production and nonsupervisory workers: line workers, nurses, drivers, retail and restaurant staff, roughly four in five private-sector jobs. That payroll pool grows only when some combination of more people working, more hours, and higher pay is true. Subtract the inflation those workers actually face and you have the Main Street Indicator: the year-over-year growth of Main Street's real paycheck pool. Above zero, Main Street is gaining ground. Below zero, it is losing ground, whatever GDP says.
GDP can grow while paychecks do not. Headline growth with a falling labor share is exactly the mix that reads as prosperity in the data and feels like decline on the ground. This indicator makes that divergence visible one jobs report at a time.
BLS Index of Aggregate Weekly Payrolls of Production and Nonsupervisory Employees, Total Private (CES0500000035), percent change from year ago, minus CPI-U inflation (CPIAUCSL), percent change from year ago. Monthly from 1965. CPI rather than PCE is deliberate: it is closer to the basket households perceive. Two honest limits: the series is establishment payrolls, so gig and 1099 work is outside it, and it is an aggregate, so a boom at the 90th wage percentile can mask erosion below. Both biases run the same direction; if anything the indicator flatters the trend.
Updated monthly with the BLS Employment Situation release. Source data: U.S. Bureau of Labor Statistics via FRED (St. Louis Fed). Cite as: Goren, A., The Main Street Indicator, amirgoren.org/indicators.
A test of whether AI-exposed sectors are expanding profit margins by suppressing payroll relative to value added (the extraction signature) or sharing the gains with labor. It is a joint test: a margin signal and a labor signal have to move together before anything counts as displacement. Each AI-exposed sector is measured against a control of non-AI-exposed industries (the rest of the private economy), a difference-in-differences, so a broad AI effect cannot hide inside the aggregate. The margin side comes from BEA value added and compensation and is annual, updated each September; the labor side comes from BLS employment and is read quarterly.
No AI-exposed sector is in the displacement zone. Displacement requires a margin signal at strength (more than two standard deviations above the control) together with employment falling behind the control, sustained, and through 2024 the two have never coincided. The two sectors that reached a signal-strength margin, professional services and health care in 2023, were both adding workers faster than the control, which is pricing, not displacement, and both faded below the margin threshold in 2024. The one sector shedding labor relative to the control, information services (thirteen straight quarters through 2026 Q1), carries a negative margin: pay per remaining worker is rising, a talent war, not extraction. The tax distortions that would bend AI toward replacing workers are measured and in place, but no broad displacement has arrived. This page is where it will show up first.
The joint test, read together. Margin signal on the horizontal, employment versus the control on the vertical, 2023 to 2024 with the year's movement. Displacement is the lower-right zone: a margin at signal strength while jobs fall below the control. It is empty.
The margin signal (difference-in-differences z-scores, each sector minus the control, normalized to its 2015-2019 gap):
Professional, scientific & technical services (AI-exposed) 2023: +2.3 2024: +0.7
Information services (AI-exposed) 2023: +0.2 2024: -0.9
Finance & insurance (AI-exposed) 2023: -0.3 2024: +0.4
Health care (cost ratchet, not AI) 2023: +2.1 2024: +0.8
Manufacturing (reference) 2023: +0.2 2024: -0.7
The margin can only be read once a year, when BEA publishes compensation by industry. The labor side is monthly, so the labor leg of the displacement signature is tracked quarterly here. Information services has held the labor-shedding leg without a break since late 2022. If its margin turns from negative to positive while employment stays below the control, that is the first displacement signal, and it will surface on this chart before the annual margin can confirm it. Nothing else is close: health care and professional services are adding workers relative to the control, and manufacturing's relative decline is the slow, non-AI kind.
An AI-exposed sector holding more than two standard deviations of margin above the control across a multi-year window while its employment stays below the control. A single-year margin spike alongside rising employment is pricing, and does not count.
For each sector the change in gross margin scaled by lagged value added is split into output-led and payroll-led parts; the payroll-led part is differenced against the same measure for a control of non-AI-exposed industries (the rest of the private economy, about 57 percent of value added) and normalized to that sector's 2015-2019 baseline. Differencing against a non-exposed control removes the economy-wide margin cycle without erasing a broad AI-exposed effect. The margin input (compensation of employees by industry) is annual, so the margin verdict is annual and matches the signal's multi-year timescale; the 2025 point arrives with BEA's September 2026 update. The labor axis (BLS employment) is monthly, read quarterly, measured the same way against the same control. The baseline rests on five annual observations, so the z-scores are directional, not a precise test.
Data: BEA GDP-by-industry (value added) and NIPA table 6.2D (compensation), through 2024; BLS Current Employment Statistics, through 2026 Q1.
Employment growth relative to the control; below the line is the shedding leg, and only information services holds it.
The Signed Error: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7107538
The Labor-Share Screen: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7107558
Amir Goren, independent researcher. Contact: goren.amir@gmail.com.