Andrews counters that the rise in manufacturing output is largely a statistical quirk. Citing the work of economist Susan Houseman, Andrews argues that these gains were largely due to impressive increases in a single subsector — computers — and more reflected qualitative improvements in the products than production efficiency gains. Once computers are stripped out of the data, a bleaker manufacturing picture emerges.
“We weren’t making more stuff with fewer people,” Andrews writes. “[W]e were making less stuff.”
The assertion is false. As Houseman herself stated in a 2016 interview, manufacturing output was about 8 percent higher than in 1997 even with the computer industry excluded. Meanwhile, manufacturing employment declined over the same period by nearly 30 percent (approximately 17 million versus 12 million). That’s certainly making more stuff — modestly more, to be sure — with significantly fewer workers.
More importantly, while Houseman’s observations about the computer sector’s outsized contributions to output are interesting, they hardly invalidate US manufacturing’s performance in recent decades — especially when compared to other countries (which would face similar data issues). Computers are not a trivial or unimportant part of the US manufacturing sector. Just as their inclusion arguably paints a somewhat skewed picture, so would their exclusion.
Indeed, what’s the limiting principle to such logic? Would eliminating sectors that, due to changing societal or technological trends (trends that have nothing to do with trade), exert a disproportionate drag on manufacturing performance provide a more accurate portrayal of the sector’s health today?
For example, from 1997 to 2018 smoking rates among US adults nearly halved. Not surprisingly, US tobacco product manufacturing also experienced a sharp (nearly 73 percent) decline in real value-added. Similarly, decreases in paper and paperboard consumption in recent decades (When’s the last time you read a physical newspaper?) correlate with a 36 percent decline in the sector’s real value-added.
Should these and other manufacturing industries that have declined or disappeared through no possible fault of trade (the trend of dematerialization, for example) be excluded from the sector to produce a more accurate sense of domestic manufacturing’s resilience — one that would cast it in an even better light? Once one begins making such data adjustments, there’s no logical end.
Other issues related to changes in the US and global manufacturing sectors also bear consideration. As has long been observed, for example, the United States is a services-based economy, with Americans devoting an increasing amount of their spending to services rather than stuff. More money — resuming a pre-pandemic trend — is going to items such as dining out and travel than new appliances (one can only have so many refrigerators and microwaves, after all).
Given this trend, continued “dematerialization,” and other countries’ growth and industrialization (mostly devoted to serving their domestic markets), even a flatlining of US manufacturing output at its record level would be respectable. Indeed, in the face of these trends and others, it’s wholly unrealistic to expect US manufacturing output to post healthy increases ad infinitum.
And yet that always seems to be the assumption that protectionists demand free traders rebut.
None of this is to argue that trade doesn’t affect US manufacturers or that it wasn’t a factor in the historic decline in US manufacturing employment (hence my original statements that it was more a story — although not entirely one — of technology than trade and that most manufacturing jobs losses were due to automation and economic development). Of course, it was. It is wildly simplistic, however, to argue — as Andrews did — that NAFTA was passed, globalization ran rampant, and 5 million manufacturing jobs went poof with an implied monocausal relationship.
Trade was a part of the employment story, but only a part. More important were productivity improvements, especially over the long term. As economist Robert Lawrence has argued, relatively faster productivity growth is the “dominant force behind the declining share of employment in manufacturing in the United States and other industrial economies.”
The steel industry offers a good example. While the number of workers in the industry declined by 79 percent (399,000 to 83,000) from 1980 to 2017 production rose by 8 percent. In contrast, only 16 percent of the decline in manufacturing employment between 2000 and 2007 has been attributed to increased imports from China.
Notably, some of Andrews’ numbers speak to productivity’s impact. While she points out that manufacturing productivity decreased between 2011 and 2022, left unsaid is that hiring in the sector during this period went up. That’s entirely consistent with the notion that productivity is a key determinant of manufacturing employment.