Manufacturing jobs are never coming back

“Jobs and factories will come roaring back into our country,” President Donald Trump promised on “Liberation Day,” as he announced tariffs that have shocked global markets and set the country on course for a recession. “We will supercharge our domestic industrial base. We will pry open foreign markets and break down foreign trade barriers, and ultimately, more production at home will mean stronger competition and lower prices for consumers.”

This has long been the key argument behind protectionist policies like Trump’s: They will bring manufacturing jobs back to America. It’s a claim popular not just on the right, but with pro-tariff Democrats and labor unions, too. Chris Deluzio, a House Democrat from western Pennsylvania (a traditional hotbed of protectionism), has urged his party to “embrace tariffs as one component of a broader industrial strategy to revitalize American manufacturing and make whole communities that have been hollowed out by decades of bad trade policy.”

It’s a false promise. Tariffs cannot “make whole” any communities that have seen manufacturing jobs depart. That’s partly because tariffs are wildly ineffective at that purpose, as we saw in Trump’s first term, when his tariffs failed to lead to any increase in manufacturing employment, while costing jobs elsewhere.

But the bigger reason is that the fall of manufacturing employment in the US was not caused primarily by changes in policy, and changes in policy cannot reverse it. What’s happening is a transition from manufacturing to services that occurs in all countries as they get richer. 

This transition happened in countries whose policies were strongly biased toward manufacturing, like Germany, just as it did in the US. 

The root cause isn’t trade negotiators selling out the working class, but the inevitable effects of rising productivity in the manufacturing sector, plus falling demand for many manufactured goods.

Indeed, manufacturing employment isn’t just falling in the US: It’s falling worldwide. That’s the essential reality that Trump, Deluzio, and other tariff-mongers refuse to understand.

Rich countries see manufacturing employment fall

For over a century, economists have observed that as national economies grow, the workforce’s composition changes. Most commonly, the economy is broken down into three broad sectors:

• Primary, which includes agriculture, fishing, and forestry;
• Secondary, which includes manufacturing and construction;
• Tertiary, which includes the services sector.

There isn’t universal agreement on the dividing lines between these; I’ve seen mining put both in primary (because it’s taking value directly from the earth, much like forestry or farming) and in secondary (because it requires advanced machinery, like manufacturing). But the broad distinction is between agriculture, manufacturing, and services.

In a recent paper, the economists Xilu Chen, Guangyu Pei, Zheng Song, and Fabrizio Zilibotti charted how employment in the primary sector (agriculture) changes as countries get richer (thanks to fellow economist Basil Halperin for pointing me to this work):

The trend is clear: The richer a country is, the fewer workers are employed in agriculture and other primary sector activities. Large-scale mechanization of farms means that we can enjoy much more plentiful food than our ancestors a century ago, with many fewer workers producing it.

What about manufacturing? Here and in mining (which the authors also put in the secondary sector), you see a kind of U-shape:

First, as countries emerge from deep poverty, the manufacturing share of employment increases. This is the process that has happened in South Korea, Taiwan, and China since the 1980s: A push toward manufacturing for export means that more and more workers move into that sector. 

But then, as countries go from middle-income to high-income and can afford more labor-saving technologies in factories, employment in the sector falls again. This is the deindustrialization process that the US and Western Europe have experienced in recent decades.

Finally, there’s the tertiary sector, or services, where the trendline is simply upward. Rich countries see more and more of their workers enter the service sector:

Trade isn’t the main thing killing manufacturing jobs

One possible interpretation of these trends is that rich countries have simply offshored sectors like agriculture and manufacturing to poorer ones. This is the diagnosis that economic populists from Donald Trump to Sen. Bernie Sanders have offered for deindustrialization: Trade competition from China, Mexico, and the like meant that manufacturing jobs shifted from high-paying union shops in the US to low-paying jobs in those countries.

Few people make this case about agriculture, for good reason. In recent decades, the US has usually exported about as much food as it imports, and 84 percent of our food is domestically produced. At the same time, US food production has grown from 3,060 calories per person per day in 1970 to 3,875 in 2022, the same period that US trade liberalized. This isn’t an industry that’s been simply shipped overseas.

Manufacturing is more complicated, but the trade-focused story is still misguided. On paper, US manufacturing output has grown at a healthy pace in recent decades — but almost all that growth is in computer products, and the numbers are very sensitive to how one adjusts for the rapidly improving quality of those products (which is important for understanding how their price has fallen). Outside of computing, productivity growth was relatively tepid. What’s more, a sizable literature finds that Chinese import competition specifically played a role in declining manufacturing employment in the 2000s.

But was it the main reason manufacturing employment fell? Probably not. Robert Z. Lawrence, an economist at Harvard focusing on trade and manufacturing, attempts in his latest book to figure out how much of the decline in manufacturing jobs was due to trade, how much was due to productivity growth (mostly automation that enabled fewer people to produce the same output), and how much was the result of slow overall economic growth, most importantly during and after the 2001 and 2008 recessions.

For overall manufacturing, the story is simple: Lawrence finds that rapidly increasing productivity explains all job loss in the US. 

For the non-computer sector (unaffected by the measurement issues mentioned above), the picture is a little more complicated. From 2000 to 2010, half the employment losses are still due to productivity growth, but slow economic growth, caused largely by two recessions, explains much of the rest. He provides two estimates: in one, trade explains a little under a quarter of the job loss in non-computer manufacturing, and in the other, it explains none. Either way, it’s not the main part of the story. 

His estimates match those from a number of other studies, using a variety of methods, attributing somewhere in the range of 0 percent to 25 percent of the decline in manufacturing jobs to trade.

Tellingly, Lawrence notes that manufacturing employment did pick up in the aftermath of the Great Recession, but only because “productivity growth in manufacturing was negligible.”

This is a crucial, and sometimes difficult to internalize, point. The trends in the charts above, showing employment by sector, also show up in data on the economy’s overall composition, and on what people spend their money on. As countries get richer, their residents spend less and less of their income on food — a phenomenon known as Engel’s law, after economist Ernst Engel.

As countries go from middle-income to rich, spending declines on manufactured goods (other than computers) too. Just as jobs shift to services, so does spending. Lawrence finds that non-computer manufacturing has fallen as a share of the US economy mostly because the “income elasticity of demand” for manufactured goods outside computers has gotten quite low. That’s technical econ-speak for “as people’s incomes rise, they spend less of their income on this product.” There’s an upper limit on how many cars and TVs and washing machines a person can buy before it stops helping them at all.

If that’s happening — if countries getting richer means that they spend less on many manufactured goods — then essentially the only way for employment manufacturing those goods to not fall is for the sector to become less productive. As a matter of arithmetic, if a sector is making up a smaller and smaller share of output, you can’t keep the hours worked the same without seeing productivity collapse.

A potential bright spot for manufacturing jobs could be computing manufacturing. Lawrence finds that, in contrast to non-computer manufacturing, people keep buying computer products at the same rate even as they get richer. But it’s also the segment of manufacturing that’s seen the fastest productivity growth, which necessarily cuts into the sector’s employment. 

Sure enough, the number of Americans employed in computer manufacturing has been constant at about 1 million since the Great Recession, meaning the sector’s share of overall employment has shrunk.

One fact more than any other underlines the predicament for countries wanting to revive manufacturing jobs: Manufacturing employment has peaked. Not US manufacturing employment, not European manufacturing employment: global manufacturing employment. 

This is a tricky thing to measure, and data tends to come with a lag, but OECD data analyzed by economist Richard Baldwin shows the total number of manufacturing jobs peaked in 2013, at around 322 million. By 2018, the total was already down to 299 million. Other analyses have confirmed that we are either near or past the peak in global manufacturing jobs.

Did the jobs go away because they were shunted offshore? Well, no. These are global figures. “The global drop thus pretty much must be due to productivity gains,” Baldwin concludes.

Goods versus jobs

In thinking about these questions, it’s really important to distinguish between manufacturing and manufacturing jobs. There are good reasons to want a strong manufacturing sector in the US and its allied nations, not least for national security reasons. China’s capacity to produce drones (among other kinds of military materiel) is vastly greater than the US’s, and you don’t have to be hawkish or anti-China to see why that might be a risky situation for the US to be in.

There are reasonable arguments to make for targeted industrial policies to try to shift manufacturing to the US or to allies. The CHIPS and Science Act, passed under President Joe Biden and currently being dismantled by the Trump administration, was a strong attempt to do this in semiconductor manufacturing.

But we should not kid ourselves that preserving a manufacturing base in the US (and in Mexico, South Korea, and other friendly nations) will come with the creation of a huge number of manufacturing jobs. 

We want manufacturing to come with rapidly increasing productivity and automation, enabling wages to rise and good prices to fall. That’s a good future. It’s just not one where lots of people are working on an assembly line.