Put the same sentence in five mouths and you get five numbers. “Every manufacturing job creates more jobs,” says the trade association, the governor’s press office, the reshoring op-ed, and the plant-tour ribbon-cutting. Then the number of extra jobs lands anywhere from 1.4 to 15, and nobody quoting it tends to say which boundary they drew or which sector they pulled it from.
The spread is the story. There is real economics under these figures, and there is a fair amount of sales pitch riding on top of it. This page traces the common values to their sources, showing where each one is solid and where it gets stretched past its scope, and why a single phrase can legitimately produce a 10x range.
Where the number comes from
A jobs multiplier comes out of input-output economics. The government version lives in the Bureau of Economic Analysis input-output accounts and its RIMS II multiplier system; the Bureau of Labor Statistics has published related employment-requirements matrices that many independent estimates lean on. The idea is simple: when demand for a sector’s output rises, jobs appear in three places.
- Direct jobs, inside the industry itself.
- Supplier jobs, in the firms that sell inputs to it. Economists call these backward linkages.
- Induced jobs, created when direct and supplier workers spend their earnings on housing, food, healthcare, and the rest. In Bivens’s EPI framework, this category also counts the public-sector jobs that taxes on those earnings support. Other models draw the induced boundary differently.
Almost every argument about the “real” multiplier is an argument about which of those buckets you count, in which industry, measured in which unit. The figures on this page count additional jobs per direct job, so bucket 1 is the “one job” on the bottom of the ratio. A supplier-only figure counts bucket 2, the jobs at suppliers. A total figure adds bucket 3, the induced effects, which are usually the biggest of the three. Adding induced effects can double or triple the count without any new supply-chain activity at all.
A note on counting, because it trips people up. “Additional jobs” and “total jobs” are not the same. EPI’s durable-manufacturing figures of 2.9 and 7.4 are additional jobs beyond the direct one. The Milken Institute’s headline “16x” is a total that includes the direct job, which works out to 15 additional. We have put every row on the same additional-jobs basis so the rows are comparable.
One more distinction sinks a lot of headlines. A jobs multiplier counts jobs per job. A dollar multiplier counts dollars of output per dollar of output. They are different measurements, and they get quoted interchangeably.
The Reliable Confidence Score
| Source or Claim | Figure | Reliable Confidence | What It Really Means |
|---|---|---|---|
| EPI / Bivens 2019, supplier jobs only, durable manufacturing | ~2.9 supplier jobs per direct job | Highwell-bounded backward linkage | The cleanest “jobs in the supply chain” number. It is an EPI calculation built from BLS Employment Requirements Matrices, BLS employment data, and BEA accounts, and it counts both materials and capital-services suppliers. Excludes induced effects. |
| EPI / Bivens 2019, total indirect, durable manufacturing | 7.4 indirect jobs per direct job | Mediumdurable-only, mostly induced | The famous “one job, seven more” figure. About 61 percent of it is induced employment, meaning household respending plus tax-supported public-sector jobs, not suppliers, and it covers durable goods only. The number is real; the way it gets quoted usually is not. |
| EPI / Scott 2015, “The Manufacturing Footprint” | 1.4+ jobs elsewhere per direct job | Mediumconservative, manufacturing-wide, dated | The low end, often cited as the floor. A manufacturing-wide indirect estimate that excludes induced effects, from a decade ago. Defensible as a conservative number, weak as “the” multiplier. |
| NAM, 2024 IMPLAN | ~5 jobs added per worker | MediumNAM calculation, includes induced effects | NAM’s current headline: every manufacturing worker is associated with about 5 added elsewhere, indirect and induced. It is a NAM calculation using 2024 IMPLAN data, and the published page does not provide enough detail to reproduce it, so it warrants the hedge. |
| NAM, dollar figure | $2.69 of total economic impact per $1.00 | Lowwrong unit for a jobs claim | NAM describes this as total economic impact per dollar spent, a dollars measure, not a jobs multiplier. It is a legitimate figure for what it measures, and it gets cross-quoted as though it counted jobs. |
| Milken Institute 2009 (DeVol et al.), electronic computer manufacturing in California | 15 additional jobs per direct job (headline: 16x total) | Lowreal, but narrow and routinely misapplied | A real input-output estimate, but for one sector, in one state, from a 2009 report built on California data through 2007, commissioned by the California Manufacturers & Technology Association. The famous “16x” counts the direct job too. It has circulated widely since. It is not a national, all-manufacturing figure and should not be quoted as one. |
What the spread is really telling you
Lay the rows next to each other and the “disagreement” mostly dissolves. The numbers are not fighting about reality. They are answering different questions and being quoted as if they answered the same one.
There is no single manufacturing jobs multiplier. There is a supplier figure near three, a total figure above seven once you add induced effects, and a single-sector state figure at fifteen, and quoting one of them as another is where the argument breaks.
Strip a figure down to supplier jobs and durable manufacturing lands near 2.9 additional jobs per direct job. Add induced effects and the same sector jumps past 7. Switch from jobs to dollars and you get $2.69. Pull the highest-linkage sector from a single state study and you get 15. All four can be cited in good faith. None of them is “the” multiplier, and a careful reader will catch anyone who treats one as another.
Why one phrase yields six different numbers
Boundary. Supplier-only versus supplier-plus-induced is the single largest swing. In EPI’s 2019 numbers, durable manufacturing shows 2.9 supplier jobs but 7.4 total once induced effects are added. Same industry, same dataset, more than double the number depending on where you draw the line.
Sector. Manufacturing is not one thing. Petroleum, steel, and motor vehicles carry heavy supply-chain linkages. Furniture and machine shops carry light ones. Quoting a steel-mill multiplier for a furniture plant is a category error wearing a statistic.
Unit. Jobs per job and dollars per dollar are different animals. NAM publishes both, around 5 jobs added and $2.69 of total economic impact. Headlines tend to blend them into one impressive sentence.
Model. EPI builds on BLS employment-requirements matrices. NAM uses IMPLAN. Economic-development agencies often use RIMS II from BEA. The models share DNA but make different assumptions about induced effects and regional leakage, so they do not line up to the decimal.
There are other figures floating in the same cluster. The Manufacturers Alliance (MAPI) has cited roughly 3.4 non-manufacturing jobs per manufacturing job, and a 1.58 supplier-only average circulates widely under the Manufacturing Institute’s name. We left both out of the table because the versions in circulation trace through secondary sources rather than a retrievable primary document. They fit the pattern either way: every “real” multiplier sits inside a range, and the range is set by the four choices above.
Reading a multiplier without getting burned
The sharpest critique of the whole exercise comes from Robert Z. Lawrence at the Peterson Institute, and it is worth holding in mind before citing any of these numbers.
His first point is about productivity. Total employment from any sale equals the value of the sale divided by labor productivity. In his illustration, holding output per hour equal across industries, $10,000 of haircuts and $10,000 of cars support about the same number of work hours, even though cars carry a multiplier of 2.5 and haircuts a multiplier of 1. The example assumes identical productivity to make the point cleanly; the lesson survives the simplification. A bigger multiplier does not mean more jobs per dollar of spending. It means the jobs are spread across more sectors.
His second point is about leakage, and it is the one that matters most for reshoring arguments. The multiplier estimates usually quoted in these debates rest on US input-output tables that treat the whole value chain as domestic. In reality, materials and components are often imported, so a multiplier read that way overstates domestic jobs. Some input-output systems can adjust for import leakage; the problem is figures interpreted as though every upstream input were made at home. Because services lean far less on imports, a dollar spent on services can support more domestic employment than the same dollar spent on manufacturing.
His third point is about classification. In-house R&D, accounting, and janitorial work count as manufacturing value added; outsource the same functions and they count as services. Part of the “manufacturing multiplier” is just a bookkeeping choice about where activity gets filed.
So, using these numbers without getting burned:
- State your boundary. “Supplier jobs only” and “including induced effects” are different claims; say which one you mean.
- Match the unit to the question. If you are talking jobs, do not reach for the $2.69 dollar figure.
- Use sector-specific multipliers when you mean a specific industry, not the headline average.
- For policy and reshoring claims, remember that these figures, as usually cited, treat the supply chain as fully domestic. When inputs are imported, the real domestic effect is smaller than the table implies.
Where writers most reliably go wrong: quoting the 7.4 figure as “every manufacturing job creates seven more” with no mention that it is durable-goods only and mostly induced; treating the 16x figure, a single California sector from a 2009 study, as a national, all-manufacturing number; and presenting a multiplier as net new domestic jobs a policy will create, which ignores both leakage and the productivity point.
How we scored this
Each row is rated on two things at once: how well-sourced the figure is, and how well it fits the claim people make with it, which is “every manufacturing job creates X more.”
A figure can be perfectly real and still rate Low here if it is the wrong tool for that sentence. The $2.69 figure is a published NAM estimate from a 2024 IMPLAN calculation; it rates Low because it answers a dollars question and gets deployed as a jobs answer. The 16x figure rates Low for a different reason: it is a real estimate, but for one sector in one state from a 2009 report built on California data through 2007, which makes it unfit as a national, all-manufacturing number. It is not comparable to EPI’s national figure for the broader computer-and-peripheral-equipment category, so neither one disproves the other.
The High rating goes to the supplier-only durable figure because it is narrowly bounded, built directly from the source data, and means exactly what it says. We also note the interests behind the largest numbers as a matter of disclosure: EPI’s 2019 paper was funded by the Alliance for American Manufacturing, NAM is an advocacy body, and the Milken Institute’s 16x estimate appeared in a report commissioned by the California Manufacturers & Technology Association. Disclosure is not an accusation. Funding does not make a number wrong, and it does not by itself tell you the number was shaded. It does tell you who paid for the study, which is worth knowing.
One data note. In February 2026, BLS removed its Input-Output matrix and related Employment Requirements Matrix tables after finding incorrect value-added percentages. That does not invalidate Bivens’s 2019 analysis, which used earlier vintages of the data, but it does mean a reader cannot currently reproduce these supplier and induced figures from live BLS tables.
If you cite just one thing
There is no single, correct manufacturing jobs multiplier, and anyone who gives you one without a boundary, a sector, and a unit has skipped the part that matters. For a defensible general figure, the honest range is roughly 1.4 to 3 supplier jobs per manufacturing job, climbing past 7 only once you fold in induced effects, meaning household respending plus public-sector jobs, and stay inside durable goods. The $2.69 figure is about dollars, not jobs. The 16x figure is real but describes one sector in one state, so it does not belong in a national argument. Cite the boundary, name the source, and the number stops being a talking point and starts being a fact.
Sources
- Josh Bivens, “Updated Employment Multipliers for the U.S. Economy,” Economic Policy Institute, January 2019. https://www.epi.org/publication/updated-employment-multipliers-for-the-u-s-economy/ (PDF: https://files.epi.org/pdf/160282.pdf)
- Robert E. Scott, “The Manufacturing Footprint and the Importance of U.S. Manufacturing Jobs,” Economic Policy Institute, January 2015. https://www.epi.org/publication/the-manufacturing-footprint-and-the-importance-of-u-s-manufacturing-jobs/
- National Association of Manufacturers, “Facts About Manufacturing” (figures from NAM calculations using 2024 IMPLAN data). https://nam.org/mfgdata/facts-about-manufacturing-expanded/
- Robert Z. Lawrence, “Does Manufacturing Have the Largest Employment ‘Multiplier’ for the Domestic Economy?” Peterson Institute for International Economics, March 2017. https://www.piie.com/blogs/realtime-economic-issues-watch/does-manufacturing-have-largest-employment-multiplier-domestic
- Ross C. DeVol, Perry Wong, et al., “Manufacturing 2.0: A More Prosperous California,” Milken Institute, June 2009 (commissioned by the California Manufacturers & Technology Association). Source of the 2.5x California figure and the 16x electronic-computer-manufacturing figure. https://ajed.assembly.ca.gov/sites/ajed.assembly.ca.gov/files/hearings/MilkenFinal20090616.pdf
- S. Bureau of Economic Analysis, RIMS II regional input-output multipliers (background on how government multipliers are produced). https://www.bea.gov/resources/methodologies/RIMSII-user-guide
- S. Bureau of Labor Statistics, Inter-industry relationships (Input-Output matrix) and Employment Requirements Matrix. Note: BLS removed these files on February 6, 2026, after identifying incorrect value-added percentages. https://www.bls.gov/emp/data/input-output-matrix.htm
