The Five Million Jobs

On October 15, 1992, Ross Perot stood on a presidential debate stage and delivered the most memorable line in trade policy history: “There will be a giant sucking sound going south.” He was warning that NAFTA would drain American manufacturing jobs to Mexico, where workers earned a fraction of U.S. wages. The line was vivid, quotable, and wrong — not because the jobs didn’t disappear, but because they didn’t disappear when or how Perot predicted.

NAFTA took effect on January 1, 1994. American manufacturing employment that month: 16.9 million. Six years later, in January 2000: 17.3 million. Not only had the giant sucking sound failed to materialize, manufacturing had actually added 400,000 jobs during NAFTA’s first six years. The late 1990s were a manufacturing boom — the dot-com era generated enormous demand for computing equipment, telecommunications hardware, and the physical infrastructure of the internet. Factories in Ohio, Texas, and North Carolina were running flat out. If NAFTA was destroying jobs, it was doing a remarkably poor job of it.

Then came 2001. In a single year — from January 2001 to January 2002 — American manufacturing lost 1.5 million jobs. Another 700,000 disappeared in 2002, and 600,000 more in 2003. By January 2004, manufacturing employment had plummeted to 14.3 million — down from 17.3 million just four years earlier. In the span of one presidential term, America lost 3 million factory jobs. This was the real catastrophe, the actual sucking sound. But it had nothing to do with NAFTA. Two events converged: the dot-com bust destroyed demand for tech hardware, and China joined the World Trade Organization in December 2001, unleashing a flood of cheap manufactured goods that devastated American factories.

The economists David Autor, David Dorn, and Gordon Hanson documented this devastation in their landmark “China Shock” research. Between 2000 and 2007, they estimated that Chinese import competition directly eliminated 2.4 million American jobs — disproportionately in furniture, textiles, apparel, electronics assembly, and toys. Communities that specialized in these industries — places like Hickory, North Carolina (furniture), Dalton, Georgia (carpets), and Galesburg, Illinois (appliances) — never recovered. The China Shock was geographically concentrated, economically devastating, and politically transformative. It created the conditions for both the Tea Party movement and Donald Trump’s 2016 campaign.

The five million jobs didn’t vanish all at once. They disappeared in three distinct waves, each with different causes and different victims.

Act I: The China Shock (2001–2004). Three million jobs lost in four years. The industries hit hardest were low-tech, labor-intensive manufacturers that had survived competition with Mexico but couldn’t survive competition with a billion Chinese workers earning $2 a day. Textiles collapsed: the number of textile mill workers in the U.S. fell from 475,000 to 225,000 in five years. Furniture production migrated to Dongguan and Shenzhen. Consumer electronics assembly essentially disappeared from American soil — the last major TV factory, a Zenith plant in Springfield, Missouri, closed in 2003. These were not NAFTA industries. Mexico didn’t make cheap electronics or clothing at scale. China did, and China’s WTO entry removed the tariff barriers that had partially protected American producers.

Act II: The Great Recession (2008–2010). Another 2.3 million jobs lost. This was different from the China Shock — it wasn’t about import competition but about demand collapse. When the housing market crashed and the financial system froze, consumer spending plummeted. Auto sales fell from 17 million to 9 million vehicles per year. GM and Chrysler went bankrupt. Industrial production dropped 17% in a single year. Every manufacturing sector contracted simultaneously — a Depression-era collapse compressed into 18 months. The recovery was painfully slow: it took until 2014 to add back a million of those lost jobs, and the remaining 1.3 million never came back at all.

Act III: The Grind (2010–2026). The recovery phase. Manufacturing slowly clawed back from its 11.4 million trough in 2010 to 12.9 million by 2019 — adding 1.5 million jobs over nine years. Then COVID hit, knocking out 600,000 jobs in a single month. The post-COVID recovery was swift but incomplete: by 2026, manufacturing sits at 12.6 million, roughly where it was in 2022. The nearshoring boom and the CHIPS Act have added jobs in semiconductors and EVs, but traditional manufacturing continues to shed positions as automation advances. Today’s factories produce twice the output of 1994 with 30% fewer workers. The robots aren’t crossing the border — they’re already inside the factory.

If NAFTA wasn’t the primary cause of manufacturing job losses, did it matter at all? The honest answer is: yes, but far less than China or automation.

The Economic Policy Institute, a labor-aligned think tank, estimated that NAFTA displaced 682,900 American jobs between 1994 and 2010 — about 43,000 per year. That’s a real number affecting real communities, particularly in auto parts, textiles, and electronics assembly along the border. But set it against the 3 million jobs lost to the China Shock in four years, or the 2.3 million lost in the Great Recession in two years, and NAFTA’s contribution looks modest. The EPI’s estimate represents roughly 14% of the total decline — significant, but not the dominant force.

The more charitable reading of NAFTA is that it redistributed jobs rather than destroying them. As we showed in Episode 5, U.S. companies invested $173 billion in Mexican factories — but those factories buy American-made components, use American-designed technology, and generate $18 billion in annual profits for American shareholders. The auto industry is the clearest example: some assembly jobs moved to Mexico, but engineering, design, and high-precision component manufacturing remained in Michigan, Ohio, and Indiana. GM employs 53,000 hourly workers in the United States today — fewer than in 1994, but at higher wages and in more productive roles.

The counterfactual question — what would have happened without NAFTA? — is unanswerable but illuminating. Without NAFTA, some manufacturing would have stayed in the U.S. longer. But much of it would have gone to China or Southeast Asia anyway, without the offsetting benefits of integrated North American supply chains. Canada and Mexico buy a third of American exports — trade that supports millions of jobs in agriculture, services, and high-end manufacturing. The net effect of NAFTA on American employment is probably close to zero: jobs lost in low-skill assembly, jobs gained in exports, engineering, and services. The problem is that the losers were concentrated in specific communities while the winners were spread across the economy.

There is a third suspect in the five million jobs case, and it may be the most important one: the robot. Between 1994 and 2024, U.S. manufacturing output doubled while employment fell 28%. American factories are producing more stuff with far fewer people. A modern auto assembly line uses 1,000 robots and 2,000 humans where the 1994 equivalent used 200 robots and 5,000 humans. A steel mill that once employed 10,000 workers now produces the same tonnage with 1,500, using electric arc furnaces and continuous casting technology that eliminated entire job categories.

The Ball State University Center for Business and Economic Research published a landmark study in 2015 that attributed 88% of manufacturing job losses between 2000 and 2010 to productivity growth (automation and technology), and only 13% to trade. The finding was controversial — trade unions and protectionist politicians disputed the methodology — but subsequent research from the MIT Task Force on the Work of the Future reached similar conclusions. The jobs that disappeared were, in many cases, not “stolen” by Mexican or Chinese workers. They were eliminated by industrial robots, CNC machines, automated warehousing systems, and software that made human labor redundant.

This matters for the NAFTA debate because automation and trade are not independent forces. The threat of foreign competition accelerated the adoption of labor-saving technology in American factories. When a plant manager in Michigan knew that a competitor could produce the same part in Mexico at one-quarter the labor cost, the rational response was not to try to match Mexican wages — it was to invest in robots that eliminated the labor cost entirely. NAFTA didn’t cause automation, but it increased the urgency. The five million jobs were lost to a three-headed monster: China, automation, and trade, working together in ways that no single policy could have prevented.

Data can describe the five million jobs. It cannot describe the five million lives. Behind every BLS statistic is a factory worker in Youngstown, Ohio, who earned $28 an hour assembling Chevrolet Cruzes until the plant closed in 2019. Behind every data point is a family in Galesburg, Illinois, where the Maytag refrigerator factory that employed 2,500 people — one-quarter of the town’s adults — shut down in 2004 and moved production to Reynosa, Mexico. Behind the trend line is a community in Newton, Iowa, where Maytag’s corporate headquarters employed 4,000 people for 80 years until Whirlpool acquired the company and relocated operations.

The geographic concentration of these losses gave them political power far beyond their numbers. Manufacturing decline devastated specific places — the Rust Belt cities of Ohio, Michigan, Pennsylvania, Wisconsin, and Indiana — while the rest of the economy thrived. Between 2000 and 2016, the U.S. economy added 18 million net new jobs, mostly in healthcare, professional services, education, and technology. But those jobs were in Nashville, Austin, Denver, and suburban corridors. They were not in Flint. They were not in Canton. They were not in the small towns of the Ohio River Valley, where deindustrialization produced not just unemployment but a cascade of social consequences: opioid addiction, population decline, falling property values, shrinking tax bases, and a deep, abiding sense that the system had abandoned them.

This is the political legacy of the five million jobs, and it matters more than the economics. Donald Trump won the 2016 presidential election by carrying Michigan (by 10,704 votes), Wisconsin (by 22,748 votes), and Pennsylvania (by 44,292 votes) — three Rust Belt states where manufacturing job losses were most concentrated and NAFTA was most viscerally despised. He promised to renegotiate NAFTA, impose tariffs on Mexico and China, and bring the factories back. He delivered on the first two promises — USMCA replaced NAFTA in 2020, and tariffs went up sharply — but the third proved impossible. As we’ll examine in Episode 8, the factories didn’t come back. The trade deficit with Mexico actually accelerated after USMCA took effect.