The idea that law shapes markets—one of LPE’s foundational premises—teaches us to look for the state’s visible hand behind our economy’s biggest winners. Over the past ten years of U.S. economic policy, one hasn’t had to look particularly hard to find it. From Operation Warp Speed to the CHIPS Act to the Inflation Reduction Act to the export license deal with Nvidia to a slew of recent equity investments in leading firms, the federal government has demonstrated that it can and will offer guaranteed profits in exchange for help achieving national economic goals. It is now time to stop asking whether we should pursue industrial policy—the United States always has—and instead advance to the more interesting question: on precisely what terms will industrial policy work best and under what conditions will it disappoint us?
One of the most important aspects of industrial policy concerns how it relates to competition. Critics of industrial policy sometimes leap to the conclusion that it is antithetical to competition, and they are half right: at its worst, industrial policy can entrench incumbents. But it need not be so. In a new working paper, I examine why the government so often ends up “picking winners” and propose legal design principles for intervening in markets in a way that benefits whole industries rather than individual firms.
I use the phrase “picking winners” to refer to the act of selecting individual firms for the privileges and responsibilities of industrial policy. Picking a winner can take the form of a fiscal subsidy, a procurement contract, an exception from regulation, or any other mechanism of advantage. Ever the reality show host, President Trump loves to pick winners. The litany of Trump’s recent deals-as-policy can be recited like a catechism: the investment in Intel, the investment and offtake agreement with MP Materials, the golden share in U.S. Steel, the equity option in Westinghouse. In each of these cases, the Trump administration identified the leading firm in an industry deemed critical (semiconductors, rare earths, steel, nuclear power) and pledged the government’s support in an ad-hoc manner not necessarily available to current or future competitors.
Trump’s heavy-handedness in picking winners has put a spotlight on the practice while simultaneously obscuring how common and mundane it is across the ordinary operation of federal programs. If you look past these well-publicized examples, you can find many more instances of spending and regulatory programs that disproportionately benefit one or two firms. The Veterans Affairs program for providing care outside of VA hospitals is managed by two contractors nationwide: Optum (part of United Health) and TriWest. The VA also purchases pharmaceuticals from a single vendor, McKesson. NASA buys three times as much launch service business from SpaceX as from any other contractor. The Commerce Department has awarded an exclusive 25-year contract to AT&T to run FirstNet, a broadband network for first responders. And Boeing receives about one-third of all loan guarantees granted by the U.S. Export-Import Bank.
Do these more routine examples warrant any of the same concerns that have been raised about Trump’s corporate anointments? One might worry, for instance, about potential government corruption. One might also worry about a lack of public oversight and accountability. These concerns are pertinent, but they are not the only troubling aspects of picking winners. That is, even when the government picks a winner using the corruption-avoiding procedures of federal procurement law, and even assuming robust oversight, these single-firm arrangements would still falter on competition grounds.
In invoking competition as a key criterion for industrial policy, I do not mean to treat it as a self-justifying shibboleth. One should be able to explain why competition serves the same aims that industrial policy does. I see three major reasons why competition is conducive to industrial policy: innovation, resilience, and dependency.
First, industrial policy often aims at encouraging technological progress in sectors deemed critical to national goals, which will tend to require innovation. There is a longstanding debate on whether more or less competition is better for innovation; the Schumpeter perspective suggests that only monopolists can afford to innovate, while the Arrow perspective argues that firms innovate in order to escape competition. Philippe Aghion and co-authors have synthesized the two perspectives under an inverted U-shaped relationship, where innovation increases with competition at low levels of competition, but then falls with increasing competition at high levels of competition. Since the sectors targeted by industrial policy will tend to be on the low-competition end of the spectrum (as these sectors only have a few capable domestic producers), we can expect competition to increase innovation. This relationship between competition and innovation is evident in analyses of Chinese industrial policy, where intergovernmental competition has sprouted many competing firms in the sectors that have subsequently achieved the highest productivity growth (e.g., electric vehicles, solar photovoltaics, wind turbines, batteries, steel).
Second, industrial policy should seek to make a critical sector more resilient against shocks, and competition is a key contributor to resilience. When a supply shock is idiosyncratic (i.e., local to one geography or specific to one production technology), then the existence of alternative suppliers straightforwardly adds resilience. Moreover, in the absence of competition, firms can take advantage of a supply crisis by exercising market power and raising prices. Picking winners therefore undermines the resilience that industrial policy aims to achieve.
Third, picking winners and foreclosing competition can make the government dependent on its chosen winner. By dependent, I mean a situation where policymakers are prevented from later employing a different firm to achieve policy goals, on account of sunk costs and switching costs. This is the dynamic known as “vendor lock-in” in contracting, or what János Kornai called the “soft budget constraint” enjoyed by firms protected by the state. Dependence leads to bailouts of “too important to fail” firms, like Lockheed in 1971 or GM and Chrysler in 2008. Not all such bailouts are financial; sometimes, the government aids its high-priority firms by exempting them from ordinarily applicable law. For example, in 2019 the Department of Justice filed an amicus brief in the Ninth Circuit arguing that the court should overturn an antitrust ruling against Qualcomm, because harming Qualcomm—the leading American firm in 5G standard-setting—could harm U.S. national security in the absence of suitable domestic competitors. The dependency problem can also be understood in democratic terms. When a firm becomes too important to fail, the public loses the option to choose a new path at any later time.
In light of these problems, it would be tempting to conclude that the government should just never pick winners. But that rule would prevent industrial policy from addressing a huge swath of critical issues. More so than cronyism, the main reason that the government picks winners is that there are often only one or two firms capable of addressing the relevant policy problem. This fact is intrinsic to the motivation for industrial policy: if there were already many domestic firms capable of producing the desired output, there would not be a compelling reason for subsidy or special treatment. Consider the set of emerging national champions handpicked for federal equity stakes under the second Trump administration. In each case, was there an obvious alternative competitor the government might have included in the deal? Intel is the only domestic firm manufacturing leading-edge logic chips; MP Materials is the only vertically integrated rare earth magnet manufacturer; Westinghouse is the only firm that has completed a large nuclear reactor in recent years; and U.S. Steel is one of two remaining vertically integrated steel producers (along with Cleveland-Cliffs). Seen in this light, the problem might not be so much in the process or terms of each deal as in the underlying market structure available for industrial policy.
Acknowledging this constraint does not require surrendering to it. Assuming, for the sake of argument, that the government should support each of these sectors, it is possible to hold two ideas at once: (1) each of these firms represents the best extant private vehicle for achieving the relevant policy goal in the short term, and (2) in the long term, it will be harmful to entrench these firms against their rivals, including rivals who do not yet exist.
This dilemma—between what is expedient in the short-term and what is prudent in the long-term—motivates the framework I call anti-entrenchment industrial policy. The framework is meant to work within the constraints of picking winners. If picking winners is inevitable, as I suspect, how might practitioners of industrial policy nonetheless avoid entrenchment? Put differently, how might they generate competition even in sectors that initially lack it? Note that this mission statement is the inverse of antitrust law, which can block consolidation but cannot conjure entrants.
In theorizing anti-entrenchment, I draw inspiration from three forebears—Alexander Hamilton, Joseph Schumpeter, and Roberto Unger—each of whom has appreciated a need for balance between monopoly-like incentives and periodic disruption. In the founding document of American industrial policy, the Report on Manufactures, Hamilton advocated a reward system for innovation, but stressed that any such bounties should only last a “moderate term.” This framework lives on in U.S. patent law. Schumpeter shunned the neoclassical ideal of perfect competition, stressing the need for temporary monopoly to “secure [space] for long-range planning.” This period of creative freedom should be followed, of course, by creative destruction which “strikes not at the margins of the profits and the outputs of the existing firms but at their foundations and their very lives.” More recently, Unger has called for novel legal and institutional arrangements designed to foster periodic upheavals in the control of social resources, such as “destabilization rights” and “rotating capital funds.” The general spirit of these so-far theoretical ideals informs my attempt to develop mechanisms for undermining entrenchment in the face of concentrated markets.
Anti-entrenchment can be achieved by four key strategies. First, industrial policy should foster entry: it should aim to increase the number of credible producers at the technological frontier. But, per the dilemma outlined above, we should assume that there will not always be enough credible entrants available in the first instance. A useful response is to structure fiscal commitments as open-ended offers available to any firm that can achieve an articulated goal. We see this approach in offtake guarantees, contracts for difference, and other offers of de-risking that can be written without a specific counterparty in mind.
Second, in order to facilitate entry at a later moment, the government should require modularity. This means encouraging state-supported firms to build modular assets to facilitate swapping in new entrants based on performance or at the end of time-limited incumbency terms. Making investments modular is a way of avoiding the sunk and switching costs that drive dependency. Modularity pairs well with multi-sourcing: the strategy of hiring multiple contractors in parallel and maintaining a credible threat to shift work from one to the other. This approach was well illustrated by the Air Force’s “Great Engine War” between General Electric and Pratt & Whitney, where F-16 jets were designed with a common engine bay that could fit an engine made by either company. The more that the government itself owns critical pieces of a project, like IP, the easier it will be to swap firms in and out.
Third, the government should avoid cross-subsidies. Picking winners often ends up entrenching not just the business unit that policymakers had in mind, but other businesses under the same ownership umbrella. For example, benefits for Intel’s manufacturing business flow to its design business, and subsidies for Occidental Petroleum’s carbon capture business flow to its core hydrocarbons business. Policymakers must be careful to wall off the intended recipient from its corporate siblings. Because internal accounting separations are notoriously difficult to enforce, this prong ultimately counsels in favor of demanding divestment and structural separation for the biggest winners of industrial policy.
Fourth, industrial policy should design for exit: when the state has no option but to work with a small set of leading firms, it should structure the arrangement to reduce policy dependence on those firms, and to reduce the firms’ dependence on the state. Reducing dependence in both directions is critical to avoiding the need for bailouts. One approach is to impose conditions that firms maintain their ability to survive without government support—this could take the form of requiring some minimum level of third-party sales, or sales growing at a rate comparable to non-subsidized rivals. The exit condition also provides a reason to prefer certain contractual forms over others. For example, term loans are less susceptible to indefinite extension than are equity stakes. The implication is not that the government should never take equity, but that when it does so, it should be especially diligent about employing every other lever in the anti-entrenchment toolkit.
Anti-entrenchment industrial policy reflects an attempt to build a bridge between two normative ideals often invoked in these pages: antimonopoly and developmentalism. Some observers treat these ideals as conflicting. From the antimonopoly standpoint, the concern is that industrial policy, even if desirable in theory, tends to disadvantage small- and medium-size firms and create large concentrated winners. From the developmentalist standpoint, the worry is that national developmental goals like electrification and vaccine development are urgent, and there may not be time to wait for market structure to improve. What I call “the dilemma of picking winners” captures the truth in both perspectives. Industrial policy will tend to involve creating and rewarding monopoly in nascent sectors, which is why it needs a deliberate strategy for disentrenching its winners. By synthesizing these two approaches to political economy, anti-entrenchment industrial policy can help us pursue Hamiltonian ends by Brandeisian means.
