The digitalization of economic life and real-world data has opened up new possibilities for the study of the economic networks, regions, and sectors that ultimately determine how economic policies play out in the real world.
Such modes of thinking will be crucial for economic policymaking in a new age of geopolitical risk.
CAMBRIDGE – In 1950, George A. Lincoln of the US Military Academy’s Department of Social Sciences published Economics of National Security: Managing America’s Resources for Defense, in which he and his colleagues distilled the lessons learned, often belatedly and painfully, from industrial mobilization during World War II. But in a second edition just four years later, Lincoln offered a “complete revision” to take account of the additional lessons learned from the “partial mobilization” for the Korean War.
The second edition provided a comprehensive, detailed analysis of what Bernard Baruch, the chair of the War Industries Board in World War I, called the “5 Ms of mobilization”: manpower, materials, money, manufacturing, and morale. But it also anticipated the emerging threats of the Cold War by identifying four distinct levels of mobilization.
The first is a full mobilization, which describes what the British, more than the Americans, underwent during WWII. The second is a limited (or partial) mobilization, such as what the United States pursued in the two years prior to Pearl Harbor and following the outbreak of the Korean War. The third is a sustained high “plateau of preparedness,” equivalent to what the US maintained, to varying degrees, throughout the Cold War. And the fourth is a demobilized “normal” peacetime economy.
In previous periods of peace dating back to the American founding, government procurement – including of munitions – represented trivial claims on the country’s resources. But the economics of national security after WWII represented a new paradigm. Each of the varying levels of mobilization was distinguished by its relative demand for resources and the legitimate need for measures to override price signals. This is where the economics of national security overlaps with and can inform the economics of a broader range of industrial policies.
In all such cases, the key challenge, according to Lincoln, is to determine “the relationship between requirements and capabilities,” which in turn depends on “the existence of adequate experience, data, and statistical information.” As he put it:
“Although it is comparatively easy to speak of converting requirements for forces into terms of end items and of end items into needed raw materials, machine tools, manpower, and facilities and of converting these factors into money, the actual detailed action of conversion from one category to another is a difficult process requiring judgment and a great deal of toil and time. The capabilities side of the equation is even more difficult. We need to know what exists in the United States economy, how it can be adjusted to a security effort and the timing of the adjustment. A security effort is dynamic, with integrated activities proceeding concurrently. A miscalculation which is small, measured in dollars or tonnage, a shortage of copper, for example, may create a major disruption.”
REDISCOVERING INDUSTRIAL POLICY
Seven decades later, this mode of thinking has become newly relevant, because it offers the necessary framework for operationalizing the “modern industrial and innovation strategy” that US President Joe Biden’s national security adviser, Jake Sullivan, presented on April 27, 2023.
According to Sullivan, the US faces multiple, mounting economic- and national-security concerns, owing to its “hollowed out” industrial base, “a new environment defined by geopolitical and security competition,” an “accelerating climate crisis and the urgent need for a just and efficient energy transition,” and “the challenge of inequality and its damage to democracy.” All these problems demand “an economic mentality that champions building.” The US must ensure that it has the “capacity to produce and innovate, and to deliver public goods like strong physical and digital infrastructure and clean energy at scale.”
In response to these challenges, the administration has implemented industrial policies such as the Infrastructure Investment and Jobs Act (2021), the CHIPS and Science Act (2022), and the Inflation Reduction Act (2022). Even taken together, these do not come close to the comprehensive mobilization of resources that won WWII on “the momentum of production,” nor do they compare even to the modest partial mobilization for the Korean War.
But they are steps in the same direction. Re-shoring America’s lost capacity for high-tech manufacturing and accelerating construction of green infrastructure will require something like a melding of Lincoln’s “partial mobilization” and the “plateau of preparedness.”
Moreover, building these capacities is bound to generate upstream bottlenecks, owing to critical dependencies that will become apparent only through (frustrated) practice. Fortunately, the new US initiatives have been deployed at a time when the idea of industrial policy is regaining some legitimacy within the economics discipline. Long derided by economists as the fruitless and wasteful attempt to “pick winners,” historic examples of effective industrial policies are now being rediscovered and evaluated with rigor.
A recent survey of the literature by economists Réka Juhász, Nathan J. Lane, and Dani Rodrik calls attention to a range of approaches that have been adopted to induce structural change in market economies by shifting incentives for competition in sectors deemed strategic (as opposed to identifying specific “national champions”). Of course, the most salient example is the US Defense Department’s sponsorship of all the technologies that combined to make the digital revolution: a hugely successful exercise in operationalizing the economics of national security.
THE MIDDLE GROUND OF ECONOMICS
Whether mobilizing for war or (re)constructing advanced manufacturing capabilities in peacetime, success turns on the functioning of complex supply chains. But this truth was long forgotten – or at least under-appreciated. Not until recent supply-chain shocks did academics, policymakers, and others start paying more attention to the complicated, barely studied “meso” (middle) domain between microeconomics and macroeconomics.
While microeconomics deals with the behavior of individual agents (firms, consumers, workers, investors), macroeconomics addresses the behavior of statistical aggregates (as represented by GDP, national income, and so forth). But the space between has largely been neglected, particularly with respect to how it serves as the dynamic context in which economic policies play out. One source of this lacuna may be the simplistic faith that markets can be trusted to deliver the most efficient solution, or at least trusted more than corruptible politicians.
The issue that has called attention to this domain has been the fragility of an economy whose structure has been optimized for efficiency. The COVID-19 pandemic exposed how the longstanding commitment to maximizing returns on capital (for the benefit of shareholders and executives) meant that minimal capital was put toward maintaining buffer stocks or redundant secondary sources that could have helped absorb supply shocks. Since the systemic benefits from these investments are positive externalities, they do not factor in individual firms’ calculations.
Moreover, ignorance of the interconnected structure of the economic system also constrained efforts to address systemic fragility. That is why there is now an urgent need to understand the economy as a complex set of production networks that evolve dynamically in response to specific demands and supply-side shocks.
A good example of such work is a landmark 2020 paper by Vasco Carvalho of the University of Cambridge and his colleagues, which traced the “propagation effects” of the 2011 Great East Japan Earthquake to identify its cumulative economic impact. They examined how the “disruption caused by the earthquake and its aftermaths propagated upstream and downstream supply chains, affecting the direct and indirect suppliers and customers of disaster-stricken firms.” And by applying a “general equilibrium model” of production networks, they were able to “estimate for the overall macroeconomic impact of the shock by taking these propagation effects into account.”
This focus on production networks opens new avenues for the economics discipline. Two great economists, going against the grain, did concern themselves with this intermediate domain. The first was the Soviet-American Nobel laureate Wassily Leontief, who constructed the first input-output table to illustrate the flow of goods from primary resources to final products.
Today, the US Bureau of Economic Analysis produces national input-output tables on an annual basis, but these are necessarily static and backward-looking. Though they report on changes in the structure of the economy, they do not provide the theoretical framework and empirical information necessary to understand how shocks are transmitted through the system, and how the economic attributes of different sectors interact dynamically.
The second mesoeconomics pioneer was the Italian economist Luigi Pasinetti, whose approach, detailed in Structural Economic Dynamics, characterized different sectors of the economy by their distinctive elasticities of demand and supply with respect to price and income, along with industry-specific rates of productivity growth. The aggregate behavior of this model economy was the dynamic result of these sectors interacting with each other. But his work was purely conceptual, lacking both the data and the relevant computational resources to be put into practice.
With the digitalization of economic life and the availability of far more computing capacity, the limitations that Pasinetti faced are receding, opening up new possibilities for mesoeconomics. In the study of supply chains, mesoeconomics promises to deliver guidelines for identifying and evaluating potential points of failure and channels of propagation, calling attention to where investments in resilience are most needed. The same mode of analysis is also relevant to the study of financial networks – as the problem of “systemically important financial institutions” after the 2008 global financial crisis showed.
Yet another application of mesoeconomics is to map the dependencies entailed by industrial-policy initiatives and their systemic consequences. Since any effort to reconstruct a high-tech manufacturing base in the US will encounter many bottlenecks, mesoeconomic models can help anticipate where enabling co-investments should be targeted.
GOING DEEPER
To understand the full potential of mesoeconomics, it helps to consider what makes it distinctive from the other branches of economics. One key feature is the focus on existing relationships between firms within and across markets, supply chains, and financial networks. These relationships have identifiable characteristics along multiple dimensions, with each participant representing a node that can be characterized by the number of links to others – the node’s “degree.”
Each of these nodes can then be weighted by the volume or value of the transactions that use it. A link to another firm connects the original one to others at second, third, and higher levels. Forward links channel the flows of goods or services from an individual firm to others; backward links map flows to firms. This reveals the architecture of the network – and of the broader network of networks – highlighting the position of each node vis-à-vis the others. One measure is a node’s relative “centrality” in the set of networks in which it participates, while another captures its relative position upstream or downstream in a supply chain.
But what do we want to know about these networks? First, we want to know how they emerge in response to market forces acting on firms. We want to know how stable and resilient they are to shocks, and whether their role in a network is small and specific, or large and systemic. We also want to know how well they serve larger social or strategic purposes, and where and how effectively state interventions may strengthen or shape particular networks.
Another question is how mesoeconomists go about this sort of analysis. Fortunately, we now have a rapidly growing body of work that integrates the usual tools of economic analysis with tools specific to the study of networks. That means every set of inter-firm relationship can be characterized mathematically as a graph, which allows for the use of concepts from graph theory (such as bottlenecks, network flows, centrality statistics, cohesiveness, modularity, and – in random graph theory – percolation and phase transitions).
This new academic work is demonstrating the potential of mesoeconomics to extend the scope of research and inform policy in new ways and in new domains. Moreover, much of the recent literature includes case studies that address the “how” of mesoeconomics by developing models that exploit increasingly available micro-data to explore the networks of relationships that the data define.
For example, in periods of stability, the quest for economic efficiency and market power drives production networks toward relationships characterized by lock-in and rent seeking, not to mention consolidation and monopolization. Costly investments in enhanced resilience or production flexibility by any one firm in an extended supply chain will create a positive externality for all direct and indirect partners, but the firm incurring the initial cost is unlikely to capture these benefits fully.
Thus, the pursuit of profit at the firm level may leave the resulting network brittle. “Given market power and market incompleteness,” writeAgostino Capponi of Columbia University, Chuan Du of the US Federal Reserve, and Nobel laureate Joseph E. Stiglitz, “one should expect markets to under-invest in resilience relative to a constrained efficient benchmark.”
This is one of a growing number of creative exercises in exploring the economics of complex production networks with the instruments of network theory. The range and diversity of this research is promising, suggesting the scale of the opportunity that mesoeconomics represents.
Consider economic history. Using the tools of mesoeconomics, Princeton University’s Ernest Liu has been able to examine the evident success of industrial policy in South Korea and China. He shows that industries targeted by the state for preferential support are strategically upstream, with strong and diverse forward links. When these industries achieved accelerated growth and increased productivity, they delivered substantial benefits across the economy’s production networks, contributing to each country’s growth “miracle.”
Others have adopted a similar approach to evaluate the potential for targeted state interventions along multilevel supply chains in a “green” economy. For example, the London School of Economics’ Philippe Aghionand his colleagues’ argument for a “sector-specific industrial policy to best address the energy transition problem” may complement, or even prove more effective than, the conventional appeal for a general tax on carbon.
In a recent paper, Liu and Song Ma show that yet another recent application of mesoeconomics is to evaluate networks of innovation, “where one sector’s past innovations may benefit other sectors’ future innovations.” They point out that while “a planner valuing long-term growth should allocate more R&D toward central sectors in the innovation network … the incentive is muted in open economies that benefit more from foreign knowledge spillovers.”
The mesoeconomic approach has also been used to explore the dynamics of inflation. For example, the University of Chicago’s Elisa Rubbo works backward from an advanced New Keynesian macroeconomic model with “multiple industries and primary factors with heterogeneous supply curves,” in order to “establish necessary and sufficient conditions for changes in relative supply and demand across industries to impact aggregate inflation.” Contrary to conventional thinking, she shows, changes in relative prices can affect the inflation index.
And like virtually all domains of economics, mesoeconomics has a financial dimension. As a recent working paper from the Bank for International Settlements explains:
“Production takes time, especially when conducted through long supply chains. Working capital in the form of inventories and receivables bridges the timing mismatch between incurring costs and receiving cash from sales. To the extent that the financing cost of working capital matters, the length of supply chains is not only a matter of the economic fundamentals (such as the production technology or trade barriers) but is also shaped by financial conditions.”
Moreover, this working-capital approach reinforces the message that extended supply chains serve to amplify micro shocks with macro consequences:
“Through this theory, we highlight a novel channel for macro fluctuations through investment in working capital, which bears a strong analogy with investment in physical capital, but which operates across groups of firms, rather than at the individual firm level. We highlight the associated repercussions of financing conditions on productivity and the volume of international trade.”
Decades after Leontief’s pioneering study of the structure of the US economy, modern input-output tables detail the complex patterns of input linkages across hundreds of industries. Delving deeper into the micro domain, it is possible to identify the supplier/customer relations of millions of firms throughout the economy. Yet one must remember that these relations are dynamic: in addition to evolving in response to endogenous forces, they are also subject to unanticipated shocks, such as the recent collapse of the Francis Scott Key Bridge in Baltimore.
Fortunately, we now know that mapping these networks with the tools of mesoeconomics can guide interventions to mitigate the consequences of such dislocations.
ECONOMIC AND NATIONAL SECURITY
In his April 2023 address, Sullivan pointed to semiconductors and minerals critical to “the clean-energy future” as strategic points where economic and national security converge:
“Consider semiconductors, which are as essential to our consumer goods today as they are to the technologies that will shape our future, from artificial intelligence to quantum computing to synthetic biology. America now manufactures only around 10 percent of the world’s semiconductors, and production – in general and especially when it comes to the most advanced chips – is geographically concentrated elsewhere. This creates a critical economic risk and a national security vulnerability.
… Or consider critical minerals – the backbone of the clean-energy future. Today, the United States produces only 4 percent of the lithium, 13 percent of the cobalt, 0 percent of the nickel, and 0 percent of the graphite required to meet current demand for electric vehicles. Meanwhile, more than 80 percent of critical minerals are processed by one country, China.”
These observations are reminiscent of Lincoln’s 70 years earlier. Reflecting on how technology had transformed the mineral base of the industrial economy in the space of just 50 years, he noted:
“In 1900, industry needed little more than the few minerals known since antiquity: coal, iron, copper, tin, lead zinc, gold and silver. But technological advances have now made some 45 metallic elements and 8000 alloys of these metals essential to modern industry. To cite one example, titanium, once an impurity in iron ore, first became useful as a substitute for lead and then became important in the construction of high-speed jet aircraft.”
Lincoln’s second mention of a newly critical material referred to germanium, which “makes possible the transistor.” With unwitting historical irony, he was writing just before silicon supplanted germanium, owing to its relative ease of processing and stability at high temperatures. But the message is the same. Whatever the profile of final demand for output – and thus the shape of those outputs – the same upstream materials are essential both to a growing economy and to its mobilization base.
Thus, access to those materials and to the technologies for processing them is of first-order significance to economic and national security. Access to lithium and cobalt is as important as access to the advanced semiconductor processing technologies mastered by TSMC in Taiwan. And TSMC is in turn dependent on the extraordinary technologies embedded in the extreme ultraviolet (“EUV”) lithography systems uniquely produced by ASML in the Netherlands.
As these examples demonstrate, mesoeconomics necessarily has an international reach. This was also true in Lincoln’s time: “The Development of Foreign Sources is vital … There is a need for actions which are advantageous to both the United States and the source country” (his emphasis). Likewise, Sullivan explains that, “Ultimately, our goal is a strong, resilient, and leading-edge techno-industrial base that the United States and its like-minded partners, established and emerging economies alike, can invest in and rely upon together.”
Of course, it is not just materials that have dual use. Success in reconstructing a high-tech manufacturing base in the US will also depend on an adequate supply of the requisite skills. The advanced machine-learning techniques collectively referred to as “AI” can augment such skills in well-defined applications. But skilled humans are also needed. In July 2023, TSMC warned that a shortage of local workers with needed expertise would delay production at its new Arizona fab. Mesoeconomics necessarily must include analyses of how labor markets respond to shifting patterns of demand for skills.
A TOOL, NOT A SILVER BULLET
The vision of mesoeconomics outlined above encompasses a dynamic map of the production and distribution of goods and services alongside the evolution of demand- and supply-side developments, all of which are conditioned on technological innovations. This might seem to offer the promise – or the threat – of comprehensive economic planning at a granular level. So, are we poised to relitigate the pre-WWII debates over the feasibility of economic planning under socialism?
No, we are not, because the reach and implications of mesoeconomics do not go so far. As with all approaches to comprehending social phenomena, mesoeconomics must humble itself before the radical uncertainty that is inherent to all the individual and collective decisions it seeks to analyze. The issue is not just that systemic shocks are random and unpredictable, or potentially born of an unanticipated regime change. It is also that the actions – routine or improvised – of all participants generate responses that alter the initial conditions that motivated those actions.
Instead, what mesoeconomics offers to policymakers is the information needed to guide targeted interventions designed either to increase the resilience of the economic system on the supply side, or to enable effective responses to legitimate extra-market demands. That information necessarily includes a mapping of economic networks to identify potential vulnerabilities and bottlenecks. Armed with such insights, those designing industrial policies will have a better chance of achieving their overlapping economic- and national-security goals.
I wish to acknowledge the extraordinary guidance and encouragement of Vasco Carvalho and Matthew Elliott of the University of Cambridge, and of Daniel Goroff of the Alfred P. Sloan Foundation.