Featured image: US Steel Works offices, Gary, Indiana. The world’s largest steel mill was opened here in 1908 by Judge Elbert Gary, head of US Steel. The company was formed when JP Morgan bought out Andrew Carnegie and others. Today the Gary Works is still the largest integrated steel mill in the USA, but far smaller than newer mills in Asia. Rockefeller’s largest refinery, begun in 1899, was not far away in Whiting, Indiana.
Author photograph, May 2026.
You can listen to the 7 minute audio version of this post here.
By Gary Hoover with assistance from ChatGPT
The media is full of politicians and reporters talking about American manufacturing, “why it went away,” and how to bring it back. However, it seems very rare for any of them to really study the situation or know the facts. As my great economics teacher, Nobel winner George Stigler said about a different aspect of the economy, “The economic role of the state has managed to hold the attention of scholars for over two centuries without arousing their curiosity.” That is, they talked and talked and they suggested, recommended, and advocated for new government policies, but they never really studied them or the effects of their proposals. What worked? What didn’t? So often legislation and policy have had a multitude of unintended consequences. I see some similarities to Stigler’s comment in today’s discussions.
We are now in the very early stages of the age of Artificial Intelligence. AI in most applications is really just the next level of “automation,” a way to do jobs faster and easier. Automation is one of the most important ways to increase productivity, which is the best way to grow an economy and the most reliable way to raise wages. The role of automation and other labor-saving technologies and processes in the evolution of modern American manufacturing cannot be overlooked.
(For my own opinion on AI and the future of jobs, see my article here.)
In the late 1970s, US manufacturing jobs totaled almost 20 million, today the number is 12.6 million, a loss of over 7 million jobs, more than a third.
At the same time, between 1975 and 2025, the US added 81 million jobs, more than 10 times as many as were lost by the factories. The latest data also indicates that the average US job pays more than the average manufacturing job, so these new jobs were not all low paying jobs.
A big underlying “plate tectonic’ of the economy continues to be the rise of services and relative decline of goods. That has had a bigger impact on what types of jobs exist than any other factor. Service workers include everyone from the President to the pastor, the airline pilot to the anthropologist, and the brain surgeon to the bodyguard. Alongside tens of millions of retail, healthcare, lodging, and foodservice employees. Those numbers include substantial numbers of managers and specialists like accountants and scientists at all levels.
For example, look at the relative share of our money spent in restaurants compared with what we spend at grocery stores. That tectonic shift has been underway for at least 90 years. The two numbers crossed only in the last decade or so as “food away from home” overtook “food at home” in terms of where society found the best value for its money.
Within the context of manufacturing jobs alone, if American workers and companies had not achieved productivity gains, manufacturing would now employ an estimated 45 million people to produce the goods we now make In America. Those gains came from all types of automation, better machinery, the use of computers, new skills, and other innovations that allowed us to get more output from our inputs. The fact we can get the same production today from 12.6 million people as would have taken 45 million people fifty years ago implies that productivity gains have been responsible for over 30 million “lost manufacturing jobs.” Those numbers are challenging to estimate, but we can say with confidence that the number is very big relative to the other numbers in this story.
The impact of the international trade that politicians love to talk about (but rarely think deeply about) includes several factors at work. Cheap imports, including from China, are believed to have cost us a maximum estimate of 3 million manufacturing jobs.
Yet Americans with jobs due to imported goods rose by an estimated 20 million to about 25 million. That includes all those people working at container ports and Japanese car dealerships and many others. The number has consistently grown, growth likely to continue into the future. Those are primarily service jobs. These new jobs can also be seen as representing about a quarter of the 81 million total American job growth number over the fifty years.
In addition, US manufacturing employment due to exports added another roughly 3 million jobs, though all of that increase took place prior to 2000 and the rise of China. Some of our goods are not as globally competitive as they once were. The number of jobs in manufacturing for export has not changed much since 2000, stabilizing around 6 million jobs. The number is cyclical due to many factors including currency exchange rates, wars, epidemics, and especially the sales of airliners and other cyclical capital goods around the world.
The US also remains a large net exporter of services, accounting for more non-manufacturing jobs. Our services balance of trade is a positive $340 billion a year, supporting around 5 million jobs. Perhaps 3 million more than 50 years ago. This includes industries like accounting, management consulting, banking, and our big airlines, the largest in the world. This trade flow has been growing rapidly and serves as a partial offset to the $1.2 trillion dollar trade deficit in goods.
Keep in mind most of these trends are global. China, Germany, and Brazil know about machinery, technology, and AI, too, often with equally challenging employment transition challenges.
Your author is not unfamiliar with industrial decline. I grew up in Anderson, Indiana, a city of around 60,000 just northeast of Indianapolis in central Indiana. Anderson went from about 25,000 General Motors employees to zero in the late 20th century. So I know first hand the challenges of finding and creating the right jobs in a shifting economy, and the human toll that the economic shifts can cause.
Yet Anderson lives on and is seeing another rise due to the growth of Indianapolis in its direction. This fundamental regional dynamism could prove a more reliable source of long-term growth than the short-lived natural gas boom over a century ago which first brought industry to the region. Such a change alters the city from living by the factory whistle when I was a kid to a modern residential and service economy.
The rest of this post is a compendium of relevant data from 1975 through 2025, and in some cases the best available estimates, compiled under my direction by ChatGPT.
See what you make of this information. Think about what it means and implies for America going forward. Consider what it tells us about the jobs that are going to be needed and created in the future.
Gary Hoover
Executive Director
American Business History Center
What happened to America’s factory jobs? The honest answer is not one villain but several large forces—rising services, extraordinary productivity, domestic outsourcing, imports, and the China shock—working at different times and in different places. From 1975 to 2025:
The headline
In 1975 the United States averaged 16.9 million manufacturing jobs. Those jobs were 21.9 percent of all nonfarm payroll employment—more than one job in five. In 2025 there were 12.6 million, only 8.0 percent of a much larger job base.
That sounds like industrial collapse. It is not that simple. Total payroll jobs more than doubled, from 77.1 million to 158.4 million. Manufacturing output, measured by the Federal Reserve’s production index, rose roughly 167 percent. America makes far more with fewer people. But the human loss was real, concentrated, and often devastating to towns whose economic life revolved around a mill or plant.
A service economy—and a changed bargain
Over these fifty years, virtually all net job growth occurred in services. We wanted more health care, education, software, finance, restaurants, entertainment, logistics, and professional help. Rowthorn and Ramaswamy found the same pattern across advanced economies: rising productivity and changing demand explained most deindustrialization, with North-South trade explaining less than one-fifth in their period. This is not a worldwide disappearance: UNIDO estimated 470 million manufacturing jobs globally in 2009, about 16 percent of the workforce. Berlingieri estimates that domestic outsourcing—factories buying cleaning, security, accounting, and other work from service firms—explains about one-quarter of the measured U.S. manufacturing decline. Some of the ‘lost’ factory job was relabeled, not shipped to Shanghai.
The old wage bargain also weakened. The following chart is not adjusted for inflation, but both lines use the same dollars and the same production-and-nonsupervisory worker concept. Manufacturing paid a visible premium through much of the period. By 2025 average hourly earnings were below the all-private average. Benefits, occupation, location, and worker mix still matter, but the simple assumption that a factory job must pay more is no longer safe.
Steel: the clearest picture of automation
Steel shows why we should not confuse jobs with production. The domestic industry has faced severe import competition, yet American mills still supply most finished steel used here. The share moves with recessions, currencies, tariffs, and world overcapacity; it did not simply fall in a straight line.
Now look at Gary Works in northwest Indiana. At its height it employed about 30,000 people and could turn out on the order of eight million tons a year. Paige Williams reported in 2025 that the works could produce roughly six million tons with about 4,300 workers; a narrower Indiana University count is closer to 2,260. The exact definitions differ, but the conclusion survives: roughly three-quarters as much potential steel with perhaps one-fourteenth to one-seventh as many workers. Tons per worker increased something like five- to tenfold.
That is automation, but also better furnaces, continuous casting, computers, improved maintenance, changed product mix, and work once done inside the company now purchased outside. A modern mill needs fewer people. It also creates safer, more skilled, and often better-paid jobs. Both facts are true.
Cars: foreign company does not mean foreign-made
The automobile story corrects another common mistake. Toyota, Honda, BMW, Hyundai, Mercedes-Benz, Volkswagen, and others built large American plants. By 2023, foreign-owned multinationals employed more than three million U.S. manufacturing workers, including more than half a million in motor vehicles and parts. A Toyota assembled in Kentucky is American production even if the parent company is Japanese; a Detroit-brand vehicle assembled in Mexico is an import. Parts content makes the boundary fuzzier still.
This is why company nationality, factory location, and the origin of parts must not be collapsed into one label. They answer different questions.
Trade took jobs—and trade supports jobs
In 1975 America exported slightly more goods than it imported on the national-income measure. By 2025 goods exports were $2.1 trillion and imports $3.35 trillion. The gap is enormous, but imports are not boxes that teleport from a foreign dock to an American living room. They require ports, railroads, trucks, warehouses, wholesalers, retailers, dealerships, designers, marketers, software, finance, repair, and management.
Exports likewise support American production and supplier jobs. Commerce counts 5.7 million jobs supported by goods exports in 2023. That number includes factory workers and domestic suppliers, so it is broader than people standing on an assembly line. It is nonetheless the most consistent official measure of employment tied to exported goods.
The hidden domestic side of imports
Baughman and Francois estimated that imports supported 16 million American jobs in the earlier edition of their work and more than 21 million using 2018 data. Their measure is broad and commissioned by trade associations, so it should not be read as a government count or as proof that every import creates jobs. It does make an essential point: a Toyota dealership, a container terminal, and a warehouse full of imported shoes are American workplaces.
These jobs cannot simply be added to manufacturing jobs, and they do not erase the factory worker displaced by imports. They reveal the modern job mix: production, trade, and services are intertwined.
How much decline came from what?
There is no clean pie chart because the causes overlap. Productivity changes prices and demand. Trade encourages automation. Offshoring can move a factory step abroad while domestic outsourcing moves another step into the service column. Still, the research gives useful boundaries.
For the long pre-China period, Rowthorn and Ramaswamy put roughly two-thirds of advanced-country deindustrialization down to productivity and changing demand. For 2000–2010, Hicks and Devaraj assign 87.8 percent of lost manufacturing jobs to productivity and 13.4 percent to trade. Houseman argues that computer-sector measurement makes such productivity claims too large. Robot-specific work by Acemoglu and Restrepo finds hundreds of thousands of U.S. jobs displaced—not millions upon millions—while Autor and Salomons find that productivity can reduce jobs in the originating industry without reducing employment economy-wide.
The China shock is more tightly estimated. Caliendo, Dvorkin, and Parro find about 550,000 manufacturing jobs lost. Autor, Dorn, Hanson, and Acemoglu put direct manufacturing losses near one million and total local-market losses around two to 2.4 million. Scott and Mokhiber offer a high-end 2.8 million manufacturing estimate. For broader offshoring, Boehm, Flaaen, and Pandalai-Nayar attribute about 13 percent of the manufacturing decline they study to foreign sourcing; Lawrence and Edwards calculate roughly 2.7 million manufacturing job-equivalents associated with the 2010 trade imbalance.
Those are not interchangeable numbers. Different years, definitions, and methods explain much of the spread.
What might 2025 have looked like?
Estimates of how many manufacturing jobs we would have in the US had we not increased productivity and if we had not seen manufacturing in China become important.
What the numbers mean (ChatGPT analysis and opinion)
My best single estimates are therefore about one million more factory jobs without the China shock, about 2.4 million more without the broader movement of production abroad, and roughly five million more under a plausible world of much slower labor-saving change. But literal ‘no automation since 1975’ arithmetic yields 45 million jobs—proof that the premise changes the whole economy and should not be mistaken for a prediction.
The central lesson is neither that trade ruined America nor that automation made every loss harmless. Services rose because consumers wanted them and because companies reorganized. Automation preserved output and competitiveness while reducing headcount. Imports displaced real workers and communities, especially during the China shock, while also feeding American factories and supporting millions of American distribution and service jobs. Exports supported millions more.
America did not stop making things. It stopped needing as many people to make each thing, moved some production abroad, moved some factory functions into service companies, and created tens of millions of jobs in other fields. The national totals look adaptive. The local experience could feel like catastrophe.
A serious manufacturing policy should begin with that both-and truth: protect strategic capacity, help workers and places absorb shocks, invest in skills and modern plants, encourage exports, and remember that productivity is not the enemy. The goal cannot be to recreate 1975. It should be to create more productive American enterprises—and make sure more Americans can participate in them.
Addendum from Gemini:
U.S. services exports support roughly 4.5 to 5 million American jobs, the United States runs a substantial trade surplus in services which reached $339.5 billion in 2025, and both the overall economic size of these exports and the number of jobs they support have grown dramatically in absolute terms and as a share of the U.S. economy between 1975 and 2025. [1, 2, 3]
Current Overview of U.S. Services Trade
- Jobs Supported: Based on the International Trade Administration’s formula, every $1 billion in services exports supports roughly 4,554 jobs. With U.S. services exports totaling over $1.23 trillion in 2025, this segment directly and indirectly fuels approximately 5.6 million American jobs. [1, 2]
- Trade Balance: Unlike the persistent U.S. deficit in physical goods, the U.S. maintains a structural trade surplus in services. [4]
- Surplus Size: In 2025, the services surplus grew by 8.9% year-over-year to $339.5 billion (with $1.2349 trillion in exports versus $895.4 billion in imports). [1]
Trends and Structural Growth (1975–2025)
The trajectory of the U.S. services export sector over the last 50 years highlights a deep structural transformation: [5]
Why the Sector Continues to Grow
- The Domestic Shift: Domestically, service-providing sectors (like professional services, healthcare, and finance) expanded to represent roughly 80% of all non-farm employment in the U.S.. [5, 6]
- Global Integration: High-value knowledge exports—such as financial platforms, business services, and software intellectual property (IP)—became easily tradable across digital borders, solidifying America’s comparative global advantage. [3, 7]
Selected sources and chart notes (ChatGPT)
• BLS and Federal Reserve: Employment, earnings, and manufacturing output series
• BEA: Annual goods exports and imports
• Rowthorn and Ramaswamy: Deindustrialization—Its Causes and Implications
• UNIDO: Manufacturing and global job generation
• Berlingieri: Domestic outsourcing and the shift to services
• Autor, Dorn, Hanson, and Acemoglu: China-shock research
• Caliendo, Dvorkin, and Parro: Trade and Labor Market Dynamics
• Scott and Mokhiber: High-end China trade-deficit estimate
• Houseman: Understanding the Decline of U.S. Manufacturing Employment
• Hicks and Devaraj: The Myth and the Reality of Manufacturing in America
• Acemoglu and Restrepo: Robots and Jobs
• Autor and Salomons: Automation, productivity, and employment
• Boehm, Flaaen, and Pandalai-Nayar: Multinationals, Offshoring, and the Decline of U.S. Manufacturing
• Lawrence and Edwards: U.S. trade imbalance and manufacturing job-equivalents
• U.S. International Trade Administration: Jobs Supported by Exports
• Baughman and Francois: Imports Work for American Workers
• USGS and AISI: Steel supply and import-market-share data
• Williams and Indiana University ERI: Gary Works employment and capacity
Chart convention: solid lines are published observations or published periodic estimates. Dashed lines are models or extensions. ‘Jobs supported’ is an input-output relationship, not necessarily a count of positions that would vanish dollar-for-dollar if trade stopped.
