Here’s How the Return of American Industry Will Actually Look

Feat. the one stupid party trick you need to amaze the economists in your life…

At some point in your life, you will find yourself trapped in a conversation with an economist or some other policy expert eager to lecture you about the many nuanced truths that they understand but an unsophisticated rube never will. This is, after all, their reason for existing. Indeed, you may find yourself in the situation because you asked casually, “so, what do you do?” and your conversation partner responded, “I study the many nuanced truths of economics and public policy that the unsophisticated masses will never understand.” Resist the temptation to just walk away!

Lean in, show earnest interest, and mention a simple fact about the American economy that they will not know, do not understand, and cannot explain: worker productivity in the manufacturing sector has been falling for more than ten years.

Don’t worry, there’s no actual dispute about this fact. It’s accurate:

It’s also so outlandish, counterintuitive, and inconsistent with expectations for a reasonably functional capitalist system that your expert will not for a moment consider it might be true. Instead, you will get an eye roll and a sigh and a comment like, “oh, I think you mean employment, not productivity.” Insist you mean productivity.

“Hmm, well you might be thinking of productivity growth. [condescending smile] It is true that we’ve seen slowdowns in productivity growth.” Hold your ground. You understand the difference between levels and growth rates perfectly well. Productivity growth turned negative more than a decade ago. Its level is falling. American factories now require more workers than in 2011 to produce equivalent output. Suggest they look it up at the Bureau of Labor Statistics. Even better, show them this post (which you’ll obviously have bookmarked and at the ready).

I promise this will be fun. I did this once with the former head of the Bureau of Labor Statistics. It will work for you.

I belabor the point, and recommend it for your next cocktail hour, not only for the enjoyment of awkward social interaction, but also because the statistic should be the start rather than the end of a much longer conversation about how we think about labor productivity and what it means for the economic challenges we have faced and the economic opportunities in front of us. The narrative goes that automation has been reducing employment, especially in manufacturing, and we just need better mechanisms for coping with that reality. But that’s exactly backward. Our problem is that we have been experiencing too little productivity growth, especially for the typical worker, including from automation. Only with incentives for, and investment in, much more of it will we get the nation back on track.

---

Here, try this:

Suppose a widget factory introduced a new technology called a Worker Augmentation Module (WAM) that cuts in half the time required to make a widget. Whereas the factory had once needed two full shifts to fulfill its contracted orders for 10,000 widgets per day, now it only needs one shift. How would we describe the WAM’s effect, and would we expect workers to celebrate it?

Many of you, I suspect, are picturing the WAM as a large piece of equipment that the widgets pass through, or perhaps a robotic arm that rapidly attaches various components. In that case, you would describe the WAM as “automation.” You might note that it had rendered half the workers superfluous, likely leading to a substantial reduction in employment—sounds like the typical story of the sector’s evolution, in microcosm.

So let’s tweak the story slightly, or at least our understanding of it. The WAM is not a physical object at all, it is an app on each worker’s phone that optimizes the sequencing of necessary customizations for each individual widget as it moves through production. Is this still “automation,” or has it become “process improvement”? Change the app’s function; it no longer optimizes the process, it instead delivers real-time coaching to the worker, allowing him to execute the same steps with the same equipment much more efficiently. Or perhaps the module is a training exercise the team works through together for 30 minutes before hitting the shop floor. Now it’s “skill enhancement.” Does training eliminate jobs, to the detriment of workers? Good luck finding someone who talks about it that way.

Another adjustment to the narrative might occur to you, especially if we are talking about training rather than automation. A factory that had automated half the steps in the production of 10,000 widgets probably sounds like one poised to lay off half its workers. But a factory that had trained its workers to produce widgets twice as fast sounds like one poised to double its output and sales to 20,000 widgets daily.

But the way that productivity growth affects employment isn’t a function of the growth’s source—training and automation both mean more output from a given amount of labor. The question is what happens to output. Look at this:

This is an extremely important picture, from American Compass’s Guide to the Labor Supply. What you’re looking at is the evolution of the manufacturing sector across three eras: the post-war “golden age” from 1947–72, the downshift from 1972–2000, and the era of rapid globalization from 2000–22. In each case the red bar shows the annual rate of productivity growth, the blue bar shows the annual rate of output growth.

In the 1947–72 period in the United States, something extraordinary happened. Growing at 3.4% annually, productivity more than doubled. But growing at 4.2%, output expanded even faster. So even as every worker could produce more than twice as much, even more workers were needed. Manufacturing employment rose from roughly 14 million to roughly 17 million. Or put another way, productivity gains—much of it through automation—“destroyed” more than half of the 14 million jobs that existed in 1947, yet we ended up with 17 million jobs that paid much better, too.

From 1972 to 2000, productivity growth slowed slightly on an annual basis but still more than doubled over the period. Output growth slowed more, coming in line with productivity growth. Sector-wide, we now had the scenario where workers could produce twice as many widgets, and did produce twice as many widgets. Despite productivity “destroying” more than half of the 17 million manufacturing jobs of 1972, there were still 17 million manufacturing jobs in 2000 that paid much better, too.

Now we come to the rather different period, from 2000 to 2022, when U.S. manufacturing employment collapsed as low as 11 million before recovering to between 12 and 13 million. We are told that this is the result of automation and productivity growth, but there is no surge in productivity growth. In fact, over the entire period, the annual rate has been a bit lower than in the earlier periods and, as noted, in recent years it has even turned negative. For all the talk of automation, no evidence exists that such forces have in fact led to a divergence from the long-run trend.

What did change is that output stopped growing. Over the period, the annual rate fell by more than half from the historical norm. Industrial production is only 10% higher today than in 2000, and no higher than its 2007 peak:

What’s more, this estimate is known to have a significant measurement error, because in valuing the output of electronics it essentially credits a doubling in the speed of a computer processor as a doubling of output. Thus, government statistics report that electronics output has increased almost 25,000% in recent decades, which obviously is not a useful description of the trajectory for the manufacturing of such products in the United States. Exclude that category and look just at the rest of the manufacturing sector, and industrial output has in fact fallen substantially in recent decades, all the way back down to its mid-1990s level:

This stagnation or outright decline in manufacturing output is not some natural consequence of economic progress. To the contrary, American consumption of manufactured goods has continued to surge. But whereas, historically, that rising consumption would have created more well-paying jobs in a more productive manufacturing sector for American workers, we have chosen instead in recent years to move existing production offshore and fill new demand via imports. So thoroughly have we eviscerated manufacturing that we are getting actively worse at making things.

---

I see all this as good news. No, it’s not good that catastrophically unwise economic policy has laid low American industry and destroyed millions of livelihoods. But, recognizing that we’ve done this to ourselves, and that none of it was inevitable, we can also see the way forward to a better future. The market fundamentalists uninterested in rebuilding American industry pooh-pooh the effort as a nostalgic desire to turn back the clock and restore a 1950s economy with 1950s jobs that are never coming back and weren’t so great anyway. That’s not right. The vision is better understood as an effort to extrapolate forward the trajectory that American industry would have followed if not blown off course by unfettered globalization.

If output had continued to grow steadily in the new millennium, in line with the American appetite for it (let alone the more rapidly rising global appetite), how would the manufacturing sector look today? As in past eras, it could have doubled in size by now and, with much higher levels of investment and much greater technological progress, it could have achieved much higher productivity gains. Employment might not have increased from its 2000 level, but imagine if it had merely held constant at 17 million jobs, rather than falling to 11 million, with those jobs twice as productive and paying twice as much.

That alternative also illustrates the folly of the claim that if capital-intensive, high-productivity manufacturing does return to the United States, it will be so automated that it will not offer the desired jobs. The question is not how many people an advanced American manufacturing plant employs as compared to an old-fashioned one with the same level of output. The question is how much it employs compared to one located in China.

The goal for American reindustrialization over the next decade should be, in effect, for the United States to “catch up” to the trend line that industry would have been on if not for the catastrophic offshoring of both jobs and expertise. This is a tall task and will require substantial investment, workforce development, trade intervention, and industrial policy. But the payoff would be dramatically higher economic output, at a higher productivity level, with millions more jobs paying much higher wages. The revitalized industrial commons would in turn foster more innovation and strengthen national security. Not only the trade deficit, but also the budget deficit would fall as the nation relied less upon the rest of the world to produce what it wanted to consume.

Importantly, the pattern of economic growth and development would spread prosperity broadly, unlike the finance- and tech-driven growth that has concentrated in narrow coastal enclaves and mostly benefited a narrow segment of workers. Advanced factories with a lower number of highly productive and well-paid workers anchoring long supply chains can buttress communities every bit as effectively as the more labor-intensive factories of yore. Certainly, they do so much better than factories that moved abroad.

No one questions that a developed economy like America’s will shift more heavily toward services as it grows. That is entirely natural, and healthy. But services occupying a greater share of economic activity does not imply that manufacturing should decline in absolute terms, or that a regional economy can possibly thrive on personal services alone. The Cato Institute’s Faces of Globalization series illustrates the point well, with the story of a Georgia town devastated by the loss of its textile industry after NAFTA and then revitalized by an advanced Kia plant twenty years later. When manufacturing returns to our shores, it won’t be the number of jobs in each plant that matters, it will be the high productivity of those jobs generating valuable output for the benefit of the nation.

- Oren