Why American Manufacturing Jobs Declined

Last updated 2 weeks ago. Our resources are updated regularly but please keep in mind that links, programs, policies, and contact information do change.

Jesse Youngblood had worked six months at the Phillips Manufacturing plant in Niles, Ohio when the company shut it down. Workers were summoned by text to a meeting and told the facility was closing. Phillips Manufacturing cited financial difficulties and market conditions for the permanent closure, which cost jobs at the Niles facility.

“Why didn’t they just tell us…listen ‘It’s no longer profitable for us. We’re going to give you guys this amount of time to get your affairs in order’,” Youngblood told a local station. His union rep, Mark Murray of United Steelworkers Local 1375, said workers had asked management directly about layoffs and were told none were planned.

Stories like this get filed under a single tidy word: decline. But the decline of American manufacturing employment is not one story. There are at least three, running on top of each other.

Here is the short answer. Manufacturing jobs fell because trade, automation, and a stack of domestic policy choices all pushed employment downward, at different times and in different places. They are not rival theories competing for the same crime. They are overlapping forces, and the honest version of events refuses to crown a single culprit.

The United States still makes an enormous amount of stuff. It now makes it with far fewer people than it used to.

The Number That Peaked in 1979

Start with the headline figure, because it anchors everything else.

The Bureau of Labor Statistics reports that manufacturing employment reached an all-time peak of 19.6 million in June 1979. It has never come close since.

The more revealing number is the share. Factory work once defined the American labor market in a way that is hard to picture now.

Source: Bureau of Labor Statistics. Shares reflect the May 1953 peak, the June 1979 employment peak, and June 2019.

The postwar decades were a long climb. Between February 1961 and August 1969, manufacturing added 4 million jobs, a 27 percent jump. From July 1975 to June 1979, it added another 3 million on the way to that peak.

Then the pattern broke. The BLS notes something striking: there were more recessions (seven) in the 40 years before the peak than after it (five). Yet manufacturing failed to fully recover from any of the cyclical losses after June 1979, according to BLS, ending with about a 34 percent net loss over the following four decades.

Before 1979, factory jobs came back after every downturn. After 1979, they stopped coming back.

Output Up, Jobs Down, and a Fight Over the Data

The Bureau of Economic Analysis shows the actual value produced by private goods-making industries, which includes manufacturing (“real” here means adjusted for inflation), rising 4.5 percent in the first quarter of 2026, outpacing services. Employment, meanwhile, has flattened rather than collapsed. The June 2026 BLS jobs report described manufacturing as showing “little or no change” for the month.

The usual reading: productivity soared, so jobs had to fall. Robots and efficiency did the work that people used to do.

Susan Houseman, an economist at the W.E. Upjohn Institute, won the 2023 Society of Labor Economists Prize for work on data and measurement. She argues that almost all the apparent growth in real manufacturing output comes from one small corner: computers and electronic products. In a Minneapolis Fed interview, she put this industry at about 12 percent of manufacturing value added, estimating that stripping it out cuts manufacturing’s output growth (after adjusting for inflation) by roughly two-thirds and productivity growth by almost half.

When statisticians price computers, they adjust for quality. A chip twice as fast counts as a price cut, even if the sticker never moves.

Houseman and colleagues documented compound annual price declines from 1997 to 2011 of about 52 percent for microprocessors and 28 percent for desktop PCs. Those steep measured price drops make the official output figures balloon.

So the clean story that output kept growing everywhere, that productivity soared, and that jobs simply had to fall, is exactly what Houseman disputes. Her point is not that the data are fake. It is that the data are being misread because of how we measure ICT products and globalization.

The China Shock and the Trade Deficit

The trade story has the sharpest turning point: December 2001, when China joined the World Trade Organization.

U.S. imports from China then exploded. Economist Robert Scott of the Economic Policy Institute, in April 2017 testimony before the House Subcommittee on Asia and the Pacific, noted that imports from China rose from $102.3 billion in 2001 to $483.2 billion in 2015.

Scott attributed 3.4 million eliminated U.S. jobs between 2001 and 2015 to the growing trade deficit with China, with manufacturing accounting for a large share of the losses. The heaviest single blow landed on computer and electronic parts, which he estimated lost 1,238,300 jobs, about 36 percent of the total, including 670,800 in computer and peripheral equipment and 282,500 in semiconductors.

Scott argued the damage ran wider than the factory floor. As imports pushed workers out of industries that compete with foreign goods, he testified, those workers crowded into other jobs and dragged wages down, reducing incomes for all non-college graduates by roughly $180 billion a year.

Analysts at the Peterson Institute for International Economics, which favors open trade, offered a sharply different accounting of NAFTA, the 1994 deal with Canada and Mexico. They noted that fears the agreement would “cause a surge of unemployment” faded as the economy added almost 17 million jobs in the seven years after it took effect.

The Peterson brief, summarizing research by economist Robert Lawrence, put the trade-off in stark terms: for every net manufacturing job lost to trade with China, it estimated the U.S. economy gained about $900,000 in 2008 through cheaper goods and greater efficiency. That is cold comfort to a laid-off worker, and the brief does not pretend otherwise. The claim is about the total benefit, not about who wins and who loses.

The honest reader is left holding two true things at once: trade delivered broad consumer benefits, and it concentrated brutal losses on specific people in specific towns.

When One Robot Replaces 5.6 Workers

Automation is the explanation people reach for when they want to sound sophisticated.

A widely cited National Bureau of Economic Research working paper studied local labor markets most exposed to industrial robots between 1990 and 2007. Areas that got more robots saw measurably lower employment and wages than areas that did not.

The authors’ preferred estimate: one additional robot per thousand workers cut the share of working-age people with jobs by about 0.34 percentage points and lowered wages by roughly 0.5 percent, with each new robot displacing approximately 5.6 workers. The effects hit hardest in routine, blue-collar jobs and among workers without a college degree, precisely the people factory work had long carried into the middle class.

So both stories check out.

Houseman’s measurement critique cuts against the strongest version of the automation claim, since a chunk of measured “productivity” turns out to be price statistics for chips rather than robots on real factory floors. But nobody serious denies that machines replaced some routine labor. The argument is about proportion, and it is not settled.

The Domestic Policy Case

In this telling, U.S. policy set up costs and rewards that made cutting American jobs the sensible business move.

The tax piece is the most concrete. A Cleveland Fed working paper documented that the slice of income going to factory workers (rather than to owners) fell more than 20 percentage points between the 1950s and 2016. Over the same period, the corporate tax rate dropped from roughly 46 percent to 16 percent. The authors estimate that 30 to 60 percent of that drop in workers’ income share across countries is linked with those falling rates.

A Federal Reserve Board working paper found a lopsided pattern that supporters of this view lean on hard. Raising a state’s top corporate rate by one point reduces local employment by about 0.3 to 0.5 percent. But tax cuts mostly “fail to stimulate growth in income and employment” outside recessions. That means tax cuts padded profits without reliably adding jobs.

Then there is health care.

Average annual premiums for employer family coverage reached $26,993, with employers paying about 70 percent. That, the argument goes, is a per-hour tax on hiring American workers that foreign competitors do not pay.

Economists like William Lazonick of the University of Massachusetts Lowell have documented how legal and tax rules made buying back their own shares and paying shareholders more attractive than investing in the business. As a result, firms came to treat their U.S. workforce as an expense to trim rather than something core.

What It Looked Like in Youngstown and Lordstown

Aggregate numbers hide the mechanism that mattered most to the people living it: a specific plant, on a specific day, deciding to close.

The archetype is Youngstown, Ohio. On what locals still call Black Monday in 1977, the Campbell Works closure threw about 5,000 steelworkers out of work, and roughly 50,000 steel jobs vanished across the Mahoning Valley within five years. Workers arrived for their Monday shift on September 19, 1977 and learned the mill was closing.

Decades later the pattern repeated in the same corner of Ohio. General Motors idled its Lordstown Assembly plant after decades of production, a shutdown that, according to a Jacobin report, contributed to at least 14,000 job cuts across GM’s restructuring. The bitter detail: the Los Angeles Times reported the company had extracted substantial annual concessions from Lordstown workers before shutting the factory anyway.

Aaron Applegate, an 11-year veteran of the plant, worked his last shift with roughly 1,600 colleagues. Parts of himself, he told CBS News, were “very bitter.”

These stories carry something the charts cannot. At the old Willow Run plant in Michigan, former worker Sterling Mullins recalled that “you didn’t need a high school diploma” to get hired. What disappeared was not only a paycheck. It was a ladder from limited schooling into stable, respectable adulthood.

There is a gap here. The government and research sources that quantify this decline, the BLS, the Fed, congressional testimony, contain almost no named-worker narratives. The people who tell you what a plant closing feels like are found in local news and oral-history archives, not in the productivity tables.

Can Policy Bring the Jobs Back?

The current toolkit is large, and the early evidence is mixed enough to frustrate everyone.

Tariffs are the bluntest instrument. The Kearney Reshoring Index, summarizing manufacturing jobs announced in 2024 through reshoring and foreign investment, stayed negative despite new trade barriers, improving only from about minus 115 to minus 91, and those figures reflect announced commitments rather than jobs actually filled. In plain terms: tariffs mostly shifted buying to other foreign countries rather than pulling production home, and raised the cost of materials and parts for manufacturers in the process.

The industrial-policy approach spends money instead of raising walls. The CHIPS and Science Act, Public Law 117-167, was approved on August 9, 2022. Congress.gov describes it as a law that “provides funds to support the domestic production of semiconductors.”

Whether those dollars translate into durable payrolls is a question the data cannot yet answer.

Trade rules were rewritten too. The United States-Mexico-Canada Agreement took effect on July 1, 2020, replacing NAFTA with tighter rules on how much of a car must be made in North America and new labor provisions.

And the safety net for workers who lose jobs to trade is itself in limbo. Trade Adjustment Assistance, which helps workers hit by trade with retraining and income support, expired on June 30, 2022. Representative Linda Sánchez introduced a modernization bill on March 4, 2026, titled the Trade Adjustment Assistance Modernization Act, to renew and expand it.

Which points to the tension that outlasts every policy cycle. The same forces that shrank factory employment also left a shrunken system for helping the people it displaced. Reshoring announcements and new chip factories may rebuild production, but they tend to run on automation and skilled technicians, not on the mass of routine jobs that defined 1979.

The real test is not whether America can make more things. It already does. The test is whether any of these tools can rebuild the ladder that Sterling Mullins climbed without a diploma, and so far no one has shown that they can.

Our articles make government information more accessible. Please consult a qualified professional for financial, legal, or health advice specific to your circumstances.