What Would a Job-Based Industrial Policy Look Like?

In some ways, “industrial policy” is a triumph of marketing and branding.

With most proposals for corporate tax breaks or subsidies, there is a healthy dose of skepticism and pushback, often using of critical terms like “corporate welfare” or “trickle-down economics.” But if corporate tax breaks and subsidies can be relabeled as “industrial policy,” then a considerable share of the skepticism and pushback seems to evaporate.

However, most “industrial policy” isn’t likely to raise the number of well-paid jobs by very much, or at all. Dani Rodrik uses this insight as a starting point for what a jobs-based industrial policy might look like in “Productivism and new industrial policies: learning from the past, preparing for the future” (one of the essays in by Simone Tagliapietra and Reinhilde Veugelers, Brueghel Blueprint Series #33). Rodrik writes:

As the name suggests, productivism focuses on enhancing the productive capabilities of all segments and regions of a society. While traditional forms of social assistance, especially better access to education and healthcare, can help in this regard, connecting people with productive employment opportunities requires interventions that go beyond these. It requires improvements on the demand side of the labor market as well as the supply side. Policies must directly encourage an increase in the quantity and quality of jobs that are available for the less-educated and less-skilled members of workforce, where they choose (or can afford to) live.

In the future, most of these jobs will come not from manufacturing, but from services such as health and long-term care and retail. Even if policy succeeds in reshoring manufacturing and supply chains, the impact on employment is likely to remain limited. The experience of East Asian manufacturing superstars such as South Korea and Taiwan provides a sobering example. These two countries have managed to rapidly increase the share of manufacturing value added in GDP (at constant prices), yet they have experienced steady declines in manufacturing employment ratios.

Whatever might be in support of an industrial policy of encouraging domestic semiconductor manufacturing, it’s a very capital-intensive industry that doesn’t employs a lot of people, and especially not those with lower levels of skill. Similarly, while I’m a big supporter of support for expanded R&D, the direct effect of such support is to pay for highly-skilled researchers. In addition, as Rodrik writes:

Policies that target climate change are not a substitute for good-job policies, and vice versa. Shoring up the middle class and disseminating the benefits of technology broadly through society requires an explicit good-jobs strategy. Such a strategy would not be so fixated on competition with China; it would target services instead of manufacturing, and it would focus on incentivising worker-friendly technologies.

What might a jobs-oriented industrial mean in practice? Rodrik offers two main thoughts.

First, he suggests setting up “regional business bureaus,” which would be less about handing out subsidies to companies who relocate, and more about helping firms to solve problems. Some standard examples would be if a group of employers, defined either by industry or by geographic location, would come to the business bureau with a problem: perhaps a lack of local workers with specific skills they need, or a local zoning rule that is blocking expansion, or a problem with a local intersection that is always jammed with traffic, or lack of high-speed internet, and others. The businesses would make a commitment that if this problem was addressed, they would expand employment to average and below-average wage groups. Although the business bureaus would also be able to help arrange financing for certain business investments, the priority would be on job-creation. The business bureau would then try to work with government, with local high schools and community colleges, and with banks or financing options to address the specific problem. Ideally, the process would be step-by-step, with requests made and jobs created, thus building credibility for future requests.

Second, in thinking about how government supports new technologies by encouraging research and development, and how firms choose to invest, it should be an open topic of conversation for whether these new technologies are likely to substitute or existing workers, or to complement and increase productivity of existing workers. For example, if a firm is deciding between retrofitting an existing factory with new equipment, or moving the factory to another country and investing in really good broad-band internet, the choice of investment will matter to existing workers.

A standard dynamic of the information technology revolution as it evolved through the 1990s and early 2000s was that it tended to, in the buzzword of the time, “hollow out the middle class.” The idea was that in old-style factory, a firm needed multiple layers of middle-management to monitor workers and to collect information for top management. However, computerization reduced this need to middle-management. Instead, it increased the productivity of a smaller number top managers who could now monitor and get information on their screens.

However, drawing on the work of Daron Acemoglu and Pascual Restrepo, Rodrik argues that certain new technologies may instead increase the productivity of existing labor. For example:

AI could be used in education to create more specialised tasks for teachers, personalise instruction for students, and increase effectiveness of schooling in the process. … [I]individual students have different learning styles, which requires teaching to be adapted to their specific needs. By generating real-time information on learning and making recommendations, AI tools can enable customised, smaller-group teaching. They can also allow instruction to respond more rapidly to evolving technologies and labour-market needs. Such tools are unlikely to replace teachers; they might in fact increase the demand for teachers (as well as redefine their roles) by enhancing the return to individual or small group instruction. …

[I]n healthcare … AI tools can significantly enhance the diagnostic and treatment capabilities of nurses, physicians’ aides and other medical technicians. They can, in effect, allow “less skilled” practitioners to perform tasks that only physicians with many more years of professional education have traditionally undertaken. The same logic also applies to other areas to boost job opportunities for those without the most advanced skills. For example, AI systems already enable the drawing up of simple contracts (such as wills) and the provision of many other services without the actual involvement of lawyers. To date, such systems have replaced primarily paralegals rather than lawyers themselves, but more advanced systems could enable paralegals to perform more advanced tasks, such as document review, due diligence and document drafting (Remus and Levy, 2016). Machine learning and neural networks can enable mid-level finance professionals to do financial risk assessment, loan underwriting and fraud detection tasks that would otherwise be undertaken by more senior professionals …

[A]ugmented and virtual reality technologies in manufacturing … [can enable] humans and robots to work together in performing precision tasks (rather than the latter replacing the former). Such technologies are based on smaller, more nimble robots that also enable greater customisation of production in response to specific customer needs. … More broadly, shop floor apps augment relatively unskilled labour by allowing workers to carry out operations that more-skilled employees typically perform.

I guess the new buzzword here is “cobot,” a neologism based on “collaborative robot.” For many of us, our mental vision of a factory is a giant assembly line, producing mostly identical products. If the workers on such an assembly line are doing the same repetitive task all day, then yes, many of those jobs can be replaced by automation. But perhaps the mental vision of a future assembly line should instead include a substantial number of humans walking around, with lightweight cobots rolling after them as needed, with the human jobs being complemented by the cobots.

In short, it’s wise to be skeptical about promises that “industrial policy” will more-or-less automatically expand the number of well-paid jobs at the middle and bottom of the income distribution. If you want an expansion of well-paid jobs, that needs to be its own actual focus of policy.