But as Mark Perry of the American Enterprise Institute argued in a blog post this weekend, those promises are unlikely to come true, for a very simple reason. The vast majority of job losses in American manufacturing are due not to trade deals and offshoring but increases in automation that have made factories able to produce even more goods than before with far fewer workers. And that's a trend that's almost impossible to reverse.
Although you might think American manufacturing is a shadow of its former self, U.S. factories actually churned out more cars, trucks and automotive parts in September 2016 than ever before, Perry points out. In that month, American production of motor vehicles and parts was up 21.2 percent from January 2000. Yet the number of workers required to make those parts was down 34 percent from January 2000, as the graph below shows.
It’s undoubtedly true that some Americans have lost manufacturing jobs to trade and offshoring, especially after China’s entry into the global trading system in 2001. In an influential paper published in December, economists David Autor, David Dorn and others estimated that rising competition from Chinese imports between 1999 and 2011 cost up to 2.4 million U.S. jobs.
And not all economists agree with Perry. The left-leaning Economic Policy Institute has argued that trade is the culprit for lost manufacturing jobs, not technology. Autor has suggested that perhaps the sluggish economy and lower demand for manufactured goods has played a role.
But many economists now think that trade deals had a relatively minor effect on U.S. manufacturing employment when compared with the influence of technology. A 2015 study by economists at Ball State University estimated that 88 percent of job losses in manufacturing in recent years were due not to trade, but to productivity growth. According to their estimates, higher levels of imports than exports contributed to only 13.4 percent of American job losses in the past decade.
Looking at the state of U.S. factories, it’s easy to see why these numbers might be true. Technology has made individual workers so much more productive that plants that required thousands of workers several decades ago now operate with a few hundred. As Perry says, the United States today has about 12 million factory workers — the same number as in 1941 — yet our factories today crank out about 10 times as much.
It’s not just robot assembly lines that we’re talking about. Manufacturers today now use electronic product-tracking systems that eliminate the need for numerous workers, for example. And it’s not just in the United States. Economists Robert Lawrence and Lawrence Edwards have pointed out that all industrialized countries, even export powerhouses like Germany and the Netherlands, have shed manufacturing jobs. Automation is even affecting employment in relatively low-wage countries like China and Vietnam.
In fact, in some sectors, manufacturing is actually returning to the United States after heading offshore. But because of technological advances, the manufacturing jobs of decades before are not coming with it. The American workers that factories want to hire today are often those with enough technical training to oversee robots.
To many economists, the transition of the U.S. economy from one based on manufacturing to one based on services today looks similar to the sharp decline in farming jobs in the late 19th century and through the 20th. In 1870, nearly half of American workers held jobs related to farming. By 1980, that proportion had fallen to 2 percent, as the graph from NPR shows below.
A similar seismic shift is probably underway in manufacturing. The real question is where the United States chooses to go from here. Do we invest in education and training to prepare workers for jobs in services or advanced manufacturing? Do we encourage investments in infrastructure or technology? Do we resign ourselves to the idea that the economy is essentially different and give workers a universal basic income?
These are the kind of questions the next president may have to address.
You might also like: