Industrial policy is at the heart of the current economic discourse, propelled by major legislative acts from the Biden administration. This column presents an analysis of the ‘New Economics of Industrial Policy’, synthesising emerging literature to understand these complex policies. It highlights the broader objectives of modern industrial policy, extending beyond traditional sectoral support. Key findings suggest a generally positive impact of these policies, with nuances in implementation and efficacy. For a full efficiency evaluation of industrial policy, future theoretical work, informed by careful empirical work on a case-by-case basis, is called for.
The rise of ‘Bidenomics’ and its signature economic legislations, such as the Infrastructure Investment and Jobs Act (IIJA), Creating Helpful Incentives to Produce Semiconductors Act (CHIPS), and Inflation Reduction Act (IRA), has thrust industrial policy to the fore of economic policy discussions. With similar packages emerging across the OECD and beyond, there is renewed interest in understanding the workings of industrial policy. What exactly is an industrial policy? How do we measure it? And how do we evaluate it? An emerging literature in economics, which we call the New Economics of Industrial Policy, is making sense of this landscape with new tools and insights.
In a recent paper (Juhász et al. 2023), we attempt to clarify the discussion around these complex policies and synthesise the emerging literature. New work has made important methodological headway in understanding the basics of policy practice and in evaluating the efficacy of policies. The emerging picture, generally, paints a more positive view of industrial policy – but also highlights important nuances.
What is industrial policy?
We define industrial policies as those government policies that explicitly target the transformation of the structure of economic activity in pursuit of some public goal. Importantly, these policies are selective; they target some activities, but not others. Moreover, they are intentional in the sense that changing the structure of the economy is what they want to do. As such, industrial policy can be many things – our definition includes the targeted sectoral policies with which they are typically associated (e.g. support for steel, automobiles, shipbuilding, or semiconductors), but it also includes support for other targeted forms of intervention, such as R&D or exporting. Likewise, the goals of industrial policy may be broad. While historically, these policies were primarily aimed at facilitating structural transformation and industrialization in particular, today, goals include climate goals, the creation of ‘good jobs’, supply chain resiliency, national security, and more.
What is the rationale for intervention?
The economic rationale for industrial policy falls into three main categories: (1) market failures such as positive externalities which imply that the market will not provide enough of a positive activity (for example, modern manufacturing, green energy, good jobs); (2) coordination failures whereby a desirable activity may only be individually profitable if everyone else is also producing; and (3) the provision of activity-specific public inputs which are public goods (for example, the charging infrastructure needed for the uptake of electric vehicles).
The controversy surrounding industrial policy is often less about theoretical rationales – which are broad – and more about practicalities. Skeptics worry that the cure will be worse than the disease. There are two broad concerns: (1) information problems which prevent even a well-intentioned government from picking the correct activities to target; and (2) political capture, which implies that even if the government knows which activities to target, self-interested actors will divert the government away from those that create benefits to society at large. Both reasons create doubt about whether governments can ‘pick winners’.
We acknowledge these challenges, yet argue that the ultimate test of the effectiveness is not whether governments can ‘pick winners’ but whether they are able to ‘let losers go’. Although cutting losers ex post may be difficult, it is far less demanding than governmental omniscience in selecting winners ex ante. In this sense, we would argue that industrial policy is not that different from many other domains of public policy choice (education policies, stabilization policies, etc.) where the justifications for government intervention are well-established (human capital externalities, Keynesian ‘rigidities’) but what works is not obvious. Yet, unlike industrial policy, debates in these arenas typically focus on how to do policy well, not whether policy should be attempted.
While economists turned away from the study of industrial policy, the world kept using them. In fact, industrial policies are ubiquitous – and growing. Recent work measuring industrial policy using innovative methods (De Pippo et al. 2022, Juhász et al. 2022, Criscuolo et al. 2022) consistently finds that industrial policies in Western economies are widespread. For all these reasons, rather than trying to persuade policymakers to avoid them, economists should study them in order to inform the question of how to do industrial policy better. The New Economics of Industrial Policy is doing just that.
Evaluating industrial policies
Indeed, the need for careful work is pressing and evaluating industrial policy requires confronting some fundamental empirical issues. For instance, consider two types of governments: a rent-seeking one beholden to special interests and a technocratic one intervening to correct market failures. Rodrik (2012) shows that with observational data alone, one cannot distinguish the two types of governments. These issues, and those documented by Rodrik and Rodriguez (2001) and Lane (2020), highlight the myriad of ways observational data alone can be uninformative about policy efficacy.
In addition, the canonical empirical exercise whereby the researcher is able to extract the orthogonal, ‘accidental’ component of an industrial policy for evaluation may not fully resolve empirical problems either. The skeptic of industrial policy may argue that an evaluation of the random allocation of industrial policy misses all the practical (informational and political capture) challenges associated with implementing industrial policy in the real world. Optimists see ways to still recover useful quantities from the endogenously placed policy.
New, well-identified work deals with the tension between the search for exogenous variation and real-world relevance by isolating different layers of treatment. One layer, which we call the ‘economic mechanism’ evaluates the question of whether the justification for industrial policy is valid, ie. is the market failure for the targeted activity large? For example, Juhász (2018) evaluates the infant industry argument in 19th-century France using the disruption to trade resulting from a blockade against Britain. Although the paper does not address contemporary policy, the paper demonstrates that the infant industry can be a powerful economic mechanism in the real world. A second layer involves evaluating a narrow version of the efficacy question: Did the firms/industries/sectors promoted by policymakers respond in the intended direction? Recent work has been informative on this margin as well.
The credibility revolution has finally arrived in research on industrial policy
We review the findings from papers that use reduced-form research designs to evaluate three types of industrial policy: infant industry, public R&D, and place-based industrial policy. First, three recent papers (Juhász 2018, Hanlon 2020, Lane 2022) evaluate episodes that mimic cases of textbook infant industry in technological follower countries. Each paper finds some support that infant industry promotion led to increased activity in the targeted sector, though to different degrees. Lane’s (2022) study of the heavy and chemicals industry drive in 20th-century South Korea produces the clearest example of a country drastically shifting its comparative advantage using industrial policy tools.
Second, two new papers provide a fairly positive take on the scope for large-scale public R&D efforts to have large local and, more speculatively, aggregate effects (Gross and Sampat 2023, Kantor and Whaley 2023). These papers study canonical episodes of ‘moonshots’ in the US and show, that during times of national crisis, the US government was able to choose technologies, places and firms that delivered the desired outcomes. Moreover, while this was not the main intention of the policies, the papers also find evidence of long-lasting positive (mostly local) effects.
Third, a great deal of new work finds that place-based industrial policies (PBIPs) often lead to outcomes consistent with the intentions of policymakers in both lagging and declining regions. Historical natural experiments point to the potential for local manufacturing activity to spur local structural transformation and income gains that last generations (Mitrunen 2021, Garin and Rothbaum 2022). Work on European PBIPs for economically distressed regions (Criscuolo et al. 2019, Cingano et al. 2022) suggests that PBIPs can help with manufacturing job growth (in reality, dampening the decline). Similarly, policies targeted at lagging regions also find positive, and often long-lasting effects through the creation of self-sustaining agglomerations (La Point and Sakabe 2021, Incoronato and Lattanzio 2023, Cerrato 2023).
Learning from the East Asian Miracle
New work on industrial policy also moves the debate forward on the controversy over the role of industrial strategy and the East Asian economic miracle. The Asian miracle constitutes not only one of the most important episodes of modern economic development, but it remains the focal point of debates surrounding the efficacy and desirability of industrial policy. A string of new studies, starting with Lane (2022), turns to South Korea’s Heavy-Chemical Industry Drive (HCI). These studies find that policy promoted the growth and export development of targeted industries, both in the short and long run (Lane 2022), with considerable long-run welfare gains (Choi and Levchenko 2022), though possibly at the cost of increasing misallocation in the economy (Kim et al. 2022).
Quantitative work by Ernest Liu (2019) provides a useful guide for policymakers confronting the challenge of picking which industries to target, in an economy where market imperfections occur across linked sectors. Liu provides off-the-shelf sufficient statistics for optimal targeting, and his framework shows that, in certain settings, subsidizing upstream sectors minimizes policy mistakes. Liu shows that actual policies used in China and South Korea’s HCI correspond to his statistics, suggesting that the informational problems of policymakers may not be insurmountable.
New empirical work, led by Aghion et al. (2015), has only begun to scratch the surface of China’s more recent industrial policy. Bai et al. (2022) explores the impact of Chinese quid-pro-quo style FDI and study spillovers from foreign joint ventures to domestic firms. Relative to unrestricted FDI, they estimate that the quid-pro-quo FDI improved the quality of affiliated domestic models and raised their sales. Deep work on Chinese shipbuilding by Kalouptsidi (2018) and Barwick et al. (2019) point to the importance of policy design. Barwick et al. (2019) shows how not all policy levers were efficacious: while production subsidies and investment subsidies may have been useful, entry subsidies led to inefficiencies in the shipbuilding sector.
Conclusions
After years on the periphery in economics, industrial policy is now the subject of an emerging strand of research. Although nascent, the New Economics of Industrial Policy is providing a more productive assessment of industrial policy – one potentially up to the task of informing questions about how to do industrial policy well. This work unpacks diverse industrial policy episodes uncovering how success hinges on critical design details and economic context. While much of this new work utilises state-of-the-art reduced-form methods, we conclude by noting that these methods fall short of providing a full efficiency evaluation of industrial policy, which requires a model. Future work should move to tackling these challenges, informed by careful empirical work on a case-by-case basis.
This article first appeared on VoxEU on December 4, 2023. Reproduced with permission.
