Do Oren Cass’s Justifications for Industrial Policy Stack Up?

This article appeared on on August 15, 2019.

Oren Cass wants the U.S. government to adopt a manufacturing-focused “industrial policy.”

In a speech at the National Conservatism conference last month, the Manhattan Institute scholar explicitly repudiated the view that resources are usually best allocated by voluntary market trades between consumers and producers.

No, said Cass, “market economies do not automatically allocate resources well across sectors.” Some “vital sectors...suffer from underinvestment” as a result and, though naturally imperfect, a “sensible industrial policy” could improve on the outcomes we currently experience.

A belief that there is such widespread “market failure” to be corrected through the government thumbing the scale might sound familiar to those with knowledge of the socialist economic planning debates. Cass baulks at the idea the socialist label can be thrown at him. But he has not yet answered the central questions this analogy poses: Why is the government better placed to decide the industrial composition of the economy than the interaction of consumers and producers? And would the political system deliver an economically-reasoned industrial policy in practice?

Some industrial policy advocates rightly state that current policy is more interventionist than we would like, and replete with incentives, subsidies, and tax breaks that could be considered a de facto industrial policy for the economy already.

But Cass’s case is not merely a criticism of how current policy operates, or seeking to level the playing field. He explicitly says that markets do not allocate funds effectively, thus implying an explicit manufacturing-focused industrial strategy from government would be desirable even if today’s current distortions were eliminated.

Yet his speech gives no indication of how we might judge how well or badly resources are currently allocated across sectors, nor a measure of how we could judge whether there is indeed currently “underinvestment” within them.

The closest he gets is a throwaway line about the size of manufacturing in U.S. output (12 percent) being smaller than in Germany (23 percent) and Japan (19 percent). No evidence is presented for why these levels are optimal or even better.

Without this kind of information, how are we to judge Cass’s industrial policy prescriptions and whether they achieve his goals? Is economic efficiency his aim? Employment? Or something else?

In the absence of meaningful metrics for success, we must instead assess the likelihood of what he foresees as the social and economic benefits from a manufacturing-focused policy shift.

The Supposed Benefits of Manufacturing

Manufacturing, which he defines as making “physical things” (traditional manufacturing, resource extraction, energy production, agriculture, some construction), is said by Cass to have three major upsides compared to employment in other sectors:

  1. It “provides particularly well-paying, stable employment — especially for men with less formal education.”
  2. It “tends to deliver faster productivity growth because its processes are susceptible to technological advances that complement labor and increase output.”
  3. It (manufacturing production) is inextricably linked with innovation, and thus a more dynamic economy.

These are aggregated as justification for policy to incentivize or support manufacturing sectors through nine specific policy proposals, including everything from demands for government investment in basic science research right through to intrusive local content requirements, restrictions on foreign ownership and biasing the tax code towards this “productive” use of labor.

Stable Employment?

The claim that manufacturing “provides particularly well-paying, stable employment — especially for men with less formal education” is an eye-opening one, given that this debate arises in part due to a precipitous decline of manufacturing employment in the United States over recent decades.

Since 1987, employment in manufacturing has fallen by 33 percent, equivalent to around 5 million fewer jobs. As economists Kerwin Kofi Charles, Erik Hurst and Mariel Schwartz have outlined, “the declines [in employment] have been most pronounced among those with lower levels of accumulated schooling.” If “stable employment” that was good for those with “less formal education” was guaranteed by manufacturing, it’s unlikely that policy would be needed to revive it.

What Cass means then is that manufacturing *used to* provide stable employment for low-skilled workers in the post-war period. Industrial policy advocates imply either that policy somehow took a wrong turn from the 1980s and 1990s onwards, with “globalization” or liberalized trade, which destroyed this stability, or that with a few policy tweaks we could return to those good old days when low-skilled men had jobs for life.

There’s of course some truth in the memory of “jobs for life.” In the immediate post-war period, many people did work for one firm for their whole career. But this belief in widespread post-war manufacturing job stability is somewhat overplayed and infused with romanticism. Even to the extent it did exist, of course, that doesn’t mean an industrial policy will bring it back.

In 1963, for example, the median economy-wide measure of continuous years of association with an employer was 4.6 years; in 2018 it was 4.2 years. A fall, sure, but not dramatic.

Though absolute employment levels in manufacturing peaked in June 1979, in relative terms manufacturing has been in near continuous decline since WW2 too (see Figure 1). In 1947, 33 percent of the workforce were employed in manufacturing, but that had already fallen to 21.2 percent by the turn of 1980, and further to 8.5 percent today.

This is instructive of a trend which suggests “policy” — and hence globalization and trade — is only part of the equation. Across almost all high-income countries, as economies become richer the balance of activity taking place in “industry” falls. Figure 2 shows that in every G7 country, for example, the proportion of total employment in “industry” — mining and quarrying, manufacturing, construction, and public utilities — has fallen since 1991, albeit to different levels. “Manufacturing employment,” as Cass defines it, could only have been more “stable” on aggregate over long periods if you presume that this universally experienced structural trend could have been bucked.

It’s true that, at the level of the individual worker, manufacturing jobs tend to be more “secure” in any given year than a job in the rest of the economy. Since 1990, the rate of “job destruction” in manufacturing has averaged 4.5 percent of jobs per year, far below the 6.9 percent for the economy as whole.

But that ignores important historical context. The rate of annual job creation in manufacturing since the 1960s has been low too, and there was net job destruction in each decade from the 1960s through 2010. In fact, even in the supposed golden era of “stable employment” from the end of the second world war to 1980, the rate of job destruction in manufacturing was 6 percent per year, not massively dissimilar to the overall rate of job destruction today, and certainly higher than the current manufacturing employment destruction rate (see Figure 3).

Manufacturing also seems to fare worse than the broader economy during economic downturns, in both output terms and its net effect on jobs. Real GDP for the whole economy fell between 2007 Q4 and 2009 Q4 by 4.0 percent, for example, yet real value-added in manufacturing fell by a massive 14.9 percent. In that same period, manufacturing employment levels fell by 16.3 percent, compared with a much smaller overall decline in total nonfarm payrolls of 5.9 percent. Some security!

In short, manufacturing provides more stable employment in the sense that the likelihood of your job being lost in any given year is lower than for the rest of the economy. But the sector as a whole has been in long-term employment decline, even prior to contemporary globalization, with net job losses over long periods and a huge fall in the manufacturing share of total employment. Job creation and destruction likewise tend to be much more volatile in manufacturing and so disproportionately drive fluctuations in the broader economy. Perhaps the policy tools Cass favors could raise the output and employment share of manufacturing. Whether that would generate more secure employment sustainably seems unlikely.

Higher Labor Productivity?

The second reason Cass gives for favoring a manufacturing-biased industrial strategy is that manufacturing tends to experience higher rates of productivity growth.

The idea is simple enough: Cass rightly highlights that manufacturing activities tend to be more prone to automation or, as he puts it, “technological advances that complement labor and increase output.” Productivity growth is ultimately what drives improvements in living standards. Surely, then, if policy had ensured more resources towards the manufacturing sector then economic growth and the gains to ordinary workers would have been much stronger over the past 30 years?

Sadly, this view gets things completely backwards and shows a glaring contradiction with his first claim about employment. As my trade colleagues never tire of outlining, manufacturing output has continued rising and the real GDP share of manufacturing has remained steady despite a long-term decline in manufacturing employment. That’s precisely because historic productivity growth in manufacturing overall has been strong, causing (at least a large part of) the employment decline of the sector. Cass’s desire for “stable employment” and “high productivity growth” in manufacturing is thus a direct contradiction.

Productivity is about producing more output with less input (including labor). Automation and technological improvements reduce the number of workers needed to produce a given quantity of goods. Unless there is very responsive consumption demand, then, productivity improvements (fostered by both innovation and indeed trade competition) will tend to reduce the number of required workers. Workers that remain will also tend to be higher skilled and hence higher paid.

The economist Robert Lawrence outlines the story best. Suppose an innovation trickles through the manufacturing sector, raising the productivity of workers. This increases the supply of the good. The extent to which this feeds through to higher output or lower prices depends on the elasticity of demand — that is, the slope of the demand curve for the product.

For manufactured goods in general, demand is relatively unresponsive to price. As prices fall due to increased supply, the quantity demanded does not greatly change. Consumers instead pocket the savings and tend to spend more on services. The workers that remain in the sector, on higher wages, likewise spend relatively more on services with their higher incomes. That’s why, between 1947 and 2017, the share of consumer spending on goods fell from 62 percent to 33 percent, despite manufacturing output continuing to rise.

The broad statistics show this this trade-off between productivity and employment. Between 1980 and 2010, when manufacturing productivity growth was rapid, employment levels in manufacturing fell dramatically. Since 2010, when the manufacturing productivity performance has been near stagnant, employment in manufacturing has crept up.

You can see this at the sub-sector level too. Take the manufacturing industry for computers and electronic products. Over the past 30 years, the sector has seen a 1000 percent increase in productivity. That has meant employment even in that sector, where demand has grown massively, has fallen by a huge 47 percent since 1987.

This trade-off can be seen by looking at the manufacturing industries with the strongest employment performance too. Manufacturing industries that have bucked the trend with employment growth, rather than contraction, include “support activities for mining,” “beverages and tobacco,” and “food manufacturing.” Yet these are three of the six worst performing manufacturing in terms of labor productivity performance. “Beverages and tobacco” has actually seen a labor productivity decline since 1987.

In contrast, the top 5 manufacturing industries by productivity performance since 1987 — computer and electronic products, but also oil and gas, textile mills, primary metals and transportation — each saw reductions in employment between 21 percent and 78 percent of their 1987 workforces.

That’s not to say that there is a definitive pattern between employment and productivity at the level of individual industry. American apparel production for, example, has seen a dramatic employment collapse while labor productivity has stagnated over the last 30 years. This is presumably because demand for clothes has simply shifted to cheaper imports from less developed countries. Plastics employment has been fairly stable too, despite impressive productivity growth. This is presumably because plastics consumption has been rising.

This all highlights an obvious truth though. The only ways to get “stable employment” through a manufacturing industrial policy would be a) to avoid disruptive productivity improvements in sectors where demand is largely fixed, b) to ensure workers are always able to move into new manufacturing industries as labor-saving technologies proliferate, or c) to focus attention solely on sectors where demand is likely to continue growing.

Strategy a) would clearly worsen economic efficiency — it would be actively making the economy poorer. Both b) and c) depend on second-guessing future demand patterns and the likelihood of individual sectors enjoying productivity growth. Even if the economy could be engineered in this direction, rapid productivity growth industries would unlikely lead to a return of tons of stable blue-collar jobs for low education workers either.

Would it not be better to just follow a consumer-led policy where the US traded according to its comparative advantages, and the industrial structure of the economy adapted to changing domestic and global demands? One cannot help but feel lots of industrial policy advocates simply do not like the sorts of service sector jobs that have proliferated as consumer demands have shifted. They get incredibly defensive about accusations they desire the reshoring of monotonous low-skilled manufacturing activities. But these are exactly the sorts of jobs that would provide the relatively “stable employment” they desire.

In a world where a massive growth of the middle-class in China and beyond is expected to lead to a surge in demand for services and high value-added manufacturing, which the US specializes in, it would especially seem short-sighted to actively try to rebalance the economy to much wider manufacturing employment.

Even aside from the growth-destroying effects of the cronyism and rent-seeking Cass’s programs would facilitate, it seems weird to assume too that manufacturing productivity will necessarily be more robust than services sectors in future. Many believe we could be on the cusp of AI, driverless cars and robotics enabling rapid productivity growth even within the service sector.


Cass’s final claim is that manufacturing is inherently tied up with innovation. He does not explain what he means by this, other than saying a productive capacity is needed to “scale” ideas. Nor does he provide metrics of innovation he believes would improve under his policies.

That’s probably unsurprising — innovation is itself difficult to define and measure. Given that Cass cites Germany and Japan as “successful” examples of the sorts of economic structure he desires though, perhaps comparisons between the United States and these countries might be instructive?

Various global innovation indices produce very different results on which of these countries are more innovative, though the US does not perform consistently “worse” across them than either Germany or Japan. In the face of such subjectivity, we might judge overall economy-wide productivity as an appropriate proxy of past innovativeness.

OECD data suggests GDP per hour worked is around the same in Germany and the United States, and much higher in both than Japan. Average annual productivity growth across all three countries has been almost exactly the same since 1990 too, though over the past 10 years it has been significantly faster here in America (even as manufacturing productivity growth has slowed).

In fact, between 1980 and 2010 manufacturing productivity growth overall was stronger in the United States than Germany or Japan. It’s only since the financial crisis that U.S. productivity has completely stagnated (see Figure 4).

Cass cites China as another tentative success story too. But China’s growth is largely “catch-up,” and the country is still much, much poorer than the United States. Studies of specific components of China’s industrial strategy have found that while interventions did alter the country’s industrial composition, they led to “sizable distortions” and “increased industry fragmentation and idleness.”

But maybe Cass believes America has big successes to build on. Maybe if we just used his tools to eke out a slightly larger manufacturing sector, then the economy could be even more dynamic and innovative.

The belief that “industrial policy” could improve the dynamism of the economy also finds little support in the UK’s experience. A comprehensive evaluation by economists Tim Leunig and Stephen Broadberry of post-war policy found the creation of “national champions” had very poor results, while industrial subsidies were "an almost unmitigated failure” and “not successful in either supporting output or employment.” In particular, the government could not “successfully distinguish between sectors that were in inevitable decline and sectors with real prospects for the future.” The knowledge problem strikes again.

None of this is to say everything is perfect here in the United States, nor that policy currently isn’t interventionist in ways that currently harm manufacturing, innovation, and the broader economy. Too often, in making some of the arguments outlined above, libertarians are misinterpreted as being inherently hostile to manufacturing industries, rather than simply being neutral and not seeking to prefer them to service sectors. Reviewing regulatory and environmental rules, the poor application of property taxes that can bias against industrial activity at the local level, and trade protectionism that raises costs for importers of intermediate goods, could all make manufacturing more productive, while reversing policy biases against it.

But it’s worth noting that Leunig and Broadberry’s conclusions about what improves manufacturing performance are the precise opposite from the policies Cass wants to adopt. They say the things that helped make British manufacturing stronger and more productive were foreign direct investment, greater trade liberalisation and a market-based competition framework. Cass instead wants to “Tax foreign acquisition of U.S. assets, making U.S. goods relatively more attractive,” “Retaliate aggressively against mercantilist countries that undermine market competition,” and “Impose local content requirements in key supply chains like communications.”


Oren Cass asserts that markets cannot generally allocate resources efficiently by industry. Yet he provides no meaningful metrics to show this is the case, nor shows why his policies would deliver better outcomes. His two main claims about the benefits of a manufacturing sector — “stable employment” and “strong productivity growth” — are directly contradictory. A plethora of evidence suggests as countries’ get richer due to automation and technological improvements, they demand relatively more services, and so the industrial sector declines in employment terms.

It would hurt, not improve, general economic performance to try to create stable employment in manufacturing industries given these trends, and would be particularly foolish given the likely rising demand for high-end manufacturing and services (healthcare, education, insurance, finance, etc.) as the global middle-class develops.