Why Are There So Many Car Companies in China & Japan vs the US? Is Antitrust Part of Industrial Policy?

China has over 160 active automobile manufacturers, not just BYD; Japan has 10 major ones; The US just have GM, Stellantis, Ford, Harley-Davidson, and Tesla

In 1961, Japan’s Ministry of International Trade and Industry summoned the country’s automakers to deliver unwelcome news. The industry had too many players: ten manufacturers competing in a market that MITI believed could support two. Toyota and Nissan would absorb the rest. Honda, Mazda, Subaru, and the others would be folded into national champions capable of competing with General Motors and Ford.

Soichiro Honda refused. In a confrontation with MITI officials that became famous in Japanese business history, he declared that he had the right to manufacture automobiles, that the government couldn’t prevent new entrants while protecting incumbents, and that free competition would motivate the industry’s growth.

I had the right to manufacture automobiles, and they couldn’t enforce a law that would allow only the existing manufacturers to build them while preventing us from doing the same. We were free to do exactly what we wanted... If MITI wanted us to merge, then they should buy our shares and propose it at our shareholders’ meeting. After all, we were a public company. The government couldn’t tell me what to do.

He rushed two vehicles into production (the T360 mini-truck and S500 sports car) to establish a manufacturing record before MITI’s consolidation bill could pass. The bill was eventually abandoned.

Three decades later, China’s central government faced the same problem and reached the same conclusion. In 1989, Beijing announced its “three large, three small, two mini” policy, calling for consolidation of the auto industry around designated national champions. The country had too many manufacturers; they needed to merge into firms with the scale to compete globally.

The policy failed. Provincial governments, whose career incentives and tax revenues depended on local industry, resisted consolidation at every turn. Beijing tried again in the 2000s, and again in the 2010s. As recently as 2024-2025, an attempted merger of state-owned Changan and Dongfeng collapsed. Today China has 169 active automobile manufacturers and 129 EV brands.

The results confounded the planners. The companies MITI sought to eliminate went on to become globally competitive manufacturers. When the United States began imposing emissions standards in the 1970s, Honda and Mazda broke ranks with major carmakers and declared they could meet the new pollution standards. Their innovation, driven by their need to differentiate and compete, became the foundation of Japanese success in the American market. In China, the fragmentation that Beijing tried to prevent produced BYD, which overtook Tesla as the world’s largest EV manufacturer by revenue in 2024. Six of the world’s top ten EV sellers are now Chinese. Product development cycles have compressed to 18-24 months, compared to four or five years for Western automakers.

Honda bet that competitive pressure would strengthen Japan’s auto industry more than government-directed consolidation. Provincial officials in Guangdong and Zhejiang, protecting their local champions from Beijing’s merger plans, made the same bet without meaning to. Both were right. But why?

Conventional, maybe Clichéd, Wisdom

We are living through a revival of industrial policy. Governments worldwide are deploying subsidies, tariffs, tax credits, and direct investments to reshape their economies. In the United States, the CHIPS and Science Act, the Inflation Reduction Act, and a new wave of trade protection signal a decisive break from the laissez-faire consensus of the past four decades.

Yet alongside this revival runs a persistent assumption: that antitrust enforcement and industrial policy are opposing strategies. The logic seems obvious. Industrial policy aims to strengthen domestic firms; antitrust enforcement attacks them. Industrial policy seeks national champions; antitrust breaks them up. As Daniel Sokol put it in his 2015 paper “Tensions between Antitrust and Industrial Policy”, “sound antitrust law and policy is in tension with industrial policy.”

This framing treats the relationship as zero-sum. When American tech firms face antitrust scrutiny, defenders argue that breaking them up would gift advantages to Chinese competitors. When Boeing seeks to acquire rivals, supporters invoke the need to compete with Airbus. The assumption is always the same: competition and competitiveness are in tension.

We think this framing isn’t exact the right fit.

A Different Framework

Tim Wu, in a recent working paper titled “Antitrust and Industrial Policy: A Misunderstood Relationship”, argues that promoting competition through antitrust can be an important means of enhancing national competitiveness. Antitrust and industrial policy are complementary, not opposing.

Wu proposes a typology of government interventions that shape industry. Aid consists of direct assistance: subsidies, grants, tariffs, protective regulations. Support consists of indirect assistance: infrastructure, research funding, education systems. Suppression consists of interventions that deliberately weaken an industry, whether for environmental, health, or strategic reasons.

The fourth category is Discipline: interventions meant to improve an industry by forcing firms to improve performance or face losing market share. Antitrust enforcement is the primary mechanism of discipline. By preventing collusion, blocking anticompetitive mergers, and restraining exclusionary conduct, antitrust forces companies to compete. That competition, Wu argues, strengthens rather than weakens industries.

Wu draws an analogy to athletic training. An Olympic athlete needs both care and challenge, both resources like nutrition and coaching and worthy opponents. Athletes routinely relocate to train against more elite competition; Lionel Messi moved to Barcelona at age thirteen to face better players. “Like boxing champions who lack suitable opponents,” Wu quotes auto executive George Romney, “companies will become soft and flabby” without competitive pressure.

What does discipline mean operationally? It means feedback. Firms facing competition receive clear, fast signals about performance: lose customers, lose revenue, watch competitors demonstrate what’s possible. Protected firms don’t receive these signals, or receive them too late and too weakly. That’s why consolidated industries grow complacent while competitive industries adapt.

Wu identifies several distinct mechanisms, and they work differently. Banning collusion prevents firms from agreeing not to compete; the discipline is immediate. Merger control prevents consolidation that would reduce rivalry; the discipline is structural. Restraining exclusionary conduct (a dominant firm using its position to foreclose rivals) protects adjacent markets; the discipline operates at industry boundaries. Compulsory licensing of intellectual property reduces barriers to entry; the discipline creates room for new firms. These are different legal tools with different effects, but they share a common logic: they force firms to earn profits through performance rather than through suppressing competition.

The two broadest effects are worth emphasizing. First, firms that cannot collude or merge their way to profits must earn them through efficiency and innovation. Second, antitrust can reduce the exclusionary effects of dominant firms that control platforms or bottlenecks. Market power in one part of an industry can slow growth in adjacent industries. Antitrust deters efforts by monopolists to suppress competition, creating room for new firms and new industries to emerge.

This framework is a better fit for Honda’s little bet. MITI believed consolidation would produce strength through scale. Honda believed competition would produce strength through discipline, through the feedback that forces continuous improvement. The subsequent history of the Japanese auto industry suggests Honda had the better of the argument.

---

What’s the Theory?

Wu’s framework draws on a substantial academic literature, particularly the work of Philippe Aghion and colleagues on competition and innovation.

The Escape-Competition Effect

The foundational work is Aghion’s 2005 study, “Competition and Innovation: An Inverted-U Relationship”. Using panel data on UK firms, Aghion and colleagues found that innovation increases with competition up to a point, then decreases as markets become very fragmented. Wu cites this research; it provides theoretical grounding for why discipline works.

The key mechanism is what they call the “escape competition effect,” and it deserves careful attention because it explains why competitive discipline produces innovation. In industries where firms operate at similar capability levels (”neck-and-neck” competitors), increased competition reduces the profits available to firms that merely maintain the status quo. If you and your rival both make decent cars, neither of you earns much. But if you can pull ahead, through better quality, lower costs, or new features, you can capture market share and earn returns on your investment.

This creates powerful incentives to innovate. Firms invest in R&D not because they love technology but because they need to escape competitive pressure. The discipline of competition makes standing still untenable; the only way out is forward. In neck-and-neck industries, more competition means more innovation because more firms are running to escape.

The relationship reverses when firms are far apart in capability. A laggard competing against a dominant leader has little chance of catching up; innovation won’t help, so why bother? This explains the inverted-U: competition drives innovation up to a point, but too much disparity (or too much fragmentation) undermines the incentive. The sweet spot is industries with multiple capable competitors, each with a realistic shot at pulling ahead.

Aghion extended this work to industrial policy in a 2015 study using Chinese firm-level data. Industrial policy has larger positive effects on productivity in more competitive industries. When subsidies are spread across many firms rather than concentrated in national champions, the disciplinary effect of competition is preserved. The escape-competition logic holds: even subsidized firms innovate more when they face rivals.

Michael Porter’s (the professor that created the five forces framework and other tools) work on national competitive advantage reaches similar conclusions. Porter identifies domestic rivalry as one of four determinants of national competitiveness in his seminal article “The Competitive Advantage of Nations”. He identifies Japan’s auto industry, with its multiple major competitors, as exemplifying how intense competition in the domestic market produces firms that can compete globally. His policy prescription aligns with Wu’s framework: governments should stimulate local rivalry by limiting direct cooperation and enforcing antitrust regulations.

Dani Rodrik puts the point sharply: “Successful industrial policy is not about picking winners. It’s about letting the losers go.” Industrial policy fails when it protects incumbents from competitive pressure, allowing inefficient firms to survive on subsidies. It succeeds when competitive discipline provides feedback about which firms deserve continued support and which should exit.

Why Antitrust Works Differently

Traditional industrial policy faces well-documented political economy problems that antitrust enforcement largely avoids.

The Loser’s Paradox. Economists have documented that declining industries are systematically better at lobbying for government support than rising industries. Loss aversion leads policymakers to place greater weight on preventing job losses than enabling job gains. Established firms with organized workforces lobby more effectively than nascent industries with fragmented interests. The result is a tendency for subsidies to flow to yesterday’s industries rather than tomorrow’s.

Antitrust has different dynamics. Becoming an enforcement target is unwelcome; no industry collectively lobbies for investigation. Section 2 cases often involve large firms targeting small ones (AT&T attacking MCI, Microsoft attacking Netscape), creating adversarial processes that give smaller players a voice. Antitrust attention tends to focus on profitable, growing industries where there are monopoly rents worth protecting, not on declining industries seeking life support.

Information-Forcing Mechanisms. Blocking a merger or proving a Sherman Act violation requires presenting significant evidence of anticompetitive effects. This requirement limits arbitrary or corrupted decisions in ways that subsidy allocation does not.

Lobbying Dynamics. Unlike subsidies, antitrust enforcement does not fit George Stigler’s model of regulation that industries effectively purchase. The process is adversarial, differing from collective industry lobbying for tariffs or subsidies that benefit all incumbents.

It is important to not that Wu does not claim antitrust is immune to political influence (I mean look at the Second Trump Administration or China’s breakdown of Jack Ma’s Ant Group). But the dynamics differ from traditional industrial policy in ways that reduce public choice pathologies, though they do not eliminate them.

The Evidence

Does the theoretical case hold up empirically? We have evidence from automobiles, technology industries, and contemporary semiconductor policy.

Automobiles: A Natural Experiment

The global automobile industry offers something close to a natural experiment. Three major auto-producing nations took dramatically different approaches to industry structure.

China’s Failed Consolidation. The Chinese case, introduced above, rewards closer examination. Economists Hanming Fang, Ming Li, and Guangli Lu analyzed three million Chinese government documents to decode the country’s industrial policy system in their study “Decoding China’s Industrial Policies”. They found that Beijing’s consolidation efforts were not halfhearted; they were sustained, repeated, and backed by central authority. The policy failed anyway.

The mechanism of failure matters. Provincial officials didn’t oppose consolidation on principle; they opposed losing their auto plants. Career advancement in the Chinese system depends heavily on regional GDP growth. Local business taxes fund provincial budgets. When Beijing said “merge,” provincial governments heard “sacrifice your tax base and your promotion prospects for someone else’s national champion.” They resisted through bureaucratic delay, selective enforcement, and quiet support for local manufacturers.

The result was an unplanned experiment in competitive discipline. Chinese brands accounted for 76 percent of global electric vehicle sales through October 2024. The fragmentation that Beijing tried to eliminate produced the industry that now leads the world.

Japan’s Preserved Competition. Honda’s defiance preserved competitive pressure in the Japanese market. The industry maintained at least ten significant manufacturers for decades. Even today, under pressure from Chinese EVs, consolidation has been resisted; Honda-Nissan merger talks collapsed in February 2025.

When emissions standards created a new competitive dimension in the 1970s, the smaller players moved fastest. Their need to differentiate drove innovation that the larger, more comfortable Toyota and Nissan had less incentive to pursue.

America’s Consolidation. The United States took a different path, the path MITI had envisioned for Japan and Beijing has repeatedly sought for China.

By the 1930s, the American industry had consolidated to three major manufacturers. In 1962, General Motors alone held over 50 percent of the market. The Big Three’s combined share topped 90 percent.

George Romney had warned Congress in 1958 what unchecked concentration would mean: “There has not been enough competition in the automobile business in the United States to compel the Big Three to keep their products as modern.” He predicted they had grown too invested in large cars and were neglecting smaller, efficient vehicles, precisely the segment Japanese manufacturers would exploit.

The Case That Was Never Filed. Wu documents that the Justice Department did not simply decline to investigate GM. It prepared a case, drafted a complaint, and chose not to file.

During the Eisenhower and Kennedy administrations, Justice Department lawyers worked on a monopolization complaint against General Motors. According to the Wall Street Journal’s reporting at the time, the complaint reached 104 pages. It was a serial monopolization case that addressed more than 40 of GM’s mergers over the preceding decades. The main remedy would have been the divestiture of Chevrolet from the rest of GM.

Robert Kennedy, as Attorney General, was widely expected to file the suit. But it was still pending when Kennedy left office. The decision passed to the Johnson Administration, and according to the Wall Street Journal, President Johnson reserved for himself the final decision on whether to proceed.

The complaint was never filed. Donald Turner, then head of the Antitrust Division, told the New York Times that the decision never reached the White House; the case may have foundered on legal uncertainty about untested theories of shared monopoly or serial acquisition. Or politics may have played a role. The same period saw passage of the National Traffic and Motor Vehicle Safety Act, championed by Ralph Nader; the Johnson Administration may have decided that if it was going to confront the auto industry, safety regulation was the priority.

Whatever the reasons, the outcome is documented: enforcers prepared to break up GM and chose not to. This is not correlation. It is a policy choice whose consequences can be traced. The 104-page complaint sat in Justice Department files while the American auto industry continued on its consolidated path.

The Consequences. When Japanese small cars entered the American market in the 1970s, the Big Three’s responses were plagued by quality problems. The Chevrolet Vega suffered from oil leaks, warped cylinder heads, and engine fires. The Ford Pinto’s dangerous fuel tank design, chosen to save eleven dollars per car, led to widely-reported fires after low-speed collisions. By 2008, the Big Three’s combined market share had fallen to 44 percent, and GM and Chrysler required government bailouts.

What if the Justice Department had filed? The divested firms would have been in a position similar to Japanese competitors: smaller players with strong incentives to escape competition through innovation. The escape-competition effect would have applied. We cannot know for certain what would have followed. But the comparison with computing and telecommunications, where antitrust pressure produced innovation and industrial succession, suggests the counterfactual is plausible.

The decision not to file against GM may have been one of the most consequential non-decisions in American industrial history.

Technology Industries: Computing and Telecommunications

Back to Wu’s paper, he examines four American technology industries from the mid-twentieth century onward: aerospace, automotive, computing, and telecommunications. All were new industries dependent on evolving technologies. All became concentrated. All received government support through research funding and defense contracts.

The critical difference was antitrust treatment. Computing and telecommunications faced intense enforcement. Aerospace and automotive were granted de facto immunity.

Telecommunications: Seventy Years of Containing AT&T. The federal government, from the 1920s onward, accepted AT&T as a regulated monopoly. But antitrust enforcers repeatedly intervened to prevent AT&T’s monopoly from spreading to adjacent industries. This was not a single intervention or even a single policy. It was a multi-generational institutional commitment spanning seven decades and multiple administrations of both parties.

The pattern begins in 1913, when AT&T sold Western Union to settle an antitrust action, keeping the telephone monopolist out of telegraphs.

In the 1920s, AT&T operated the nation’s dominant radio network, refusing to transmit content from rivals. Following FTC pressure, AT&T sold what became NBC to RCA.

In the 1930s, FTC lawsuits forced AT&T out of the film industry, where it had expanded using its patents over film sound technology.

In 1956, a consent decree kept AT&T out of computing and semiconductors, leaving room for IBM, Fairchild, and Intel. The decree required AT&T to license 7,820 patents royalty-free. This compulsory licensing has been studied carefully: Martin Watzinger and colleagues found it increased follow-on innovation by 17 percent, with the gains concentrated in young and small companies. The mechanism is clear: reducing the exclusionary effects of AT&T’s patent portfolio created room for entrants who would otherwise have been blocked or taxed. Fairchild Semiconductor, founded in 1957, benefited directly; Intel, founded by Fairchild alumni in 1968, benefited indirectly. The semiconductor industry that now anchors American technology grew in space that antitrust enforcement created.

In the 1970s, the FCC’s Computer Inquiries kept AT&T at a distance from online services and data processing, maintaining separation between the telephone network and the computing applications that would eventually run on top of it.

The 1984 breakup extended this logic, dividing AT&T into eight firms. The breakup aided long-distance providers, online services, equipment manufacturers, and eventually the internet industry itself. Those industries developed “on top” of the telecommunications platform once AT&T could no longer foreclose them.

The cumulative effect was to maintain the telephone network as neutral infrastructure upon which other industries could develop. Each intervention created room for industries that later became significant. A small example illustrates the stakes: AT&T had a working prototype of an answering machine by the 1930s. It refused to deploy or allow connection of such devices to its network until federal intervention in the 1970s, a four-decade delay. The suppression wasn’t technological; it was structural. The monopolist had no incentive to enable devices that might reduce calls or create competition.

This is not to say the telecommunications interventions were costless or that American telecom is an unambiguous success story. The breakup imposed transition costs. Some argue that subsequent reconsolidation (SBC acquiring AT&T, the name surviving as a brand on a different company) suggests the interventions didn’t permanently restructure the industry. American broadband today is less competitive and more expensive than in many peer countries. Wu’s argument is not that antitrust produced optimal outcomes, but that it created room for adjacent industries that would not otherwise have developed. The internet industry grew in space that the telecommunications interventions created, whatever one thinks of the current state of telephone and broadband service.

What made the telecommunications pattern work was its persistence. One intervention might have been evaded or eroded. Seventy years of sustained pressure, across administrations, created an institutional expectation: AT&T would be kept in its lane. That expectation shaped behavior even when no case was pending.

Computing: Discipline Through Threat. In 1969, the Justice Department filed a monopolization case against IBM that would last fifteen years. The case was eventually dropped in 1982 as “without merit.” Yet its effects on industry structure were profound.

The mechanism is stranger than it first appears. IBM’s most consequential change came before the complaint was filed, not after. In 1969, anticipating the lawsuit, IBM unbundled software from hardware. The company’s general counsel had identified bundling as “a glaring violation of antitrust law” and advised preemptive action. IBM’s leadership agreed: better to unbundle voluntarily than be forced to do so by a court.

This decision created space for the independent software industry. Within a decade, software had become among the fastest-growing industries in the United States. The change was driven not by a court order but by corporate counsel’s fear of what a court might order.

The pattern continued through the 1970s. IBM’s behavior reflected ongoing antitrust caution. The company declined to acquire Apple, Microsoft, and Seagate when all were small businesses; its attorneys feared renewed scrutiny. When IBM entered the personal computer market, it used a modular design with non-exclusive supplier relationships. The IBM-PC had a hard drive from Seagate, a printer from Epson, a processor from Intel, and (most fatefully) an operating system from a startup called Microsoft. The non-exclusivity was crucial: it enabled Compaq, Dell, and others to build compatible machines.

Wu describes the result as “divided technological leadership,” an era in which no single monopoly controlled computing, but interoperating firms at different levels competed and innovated. The antitrust pressure didn’t just affect IBM’s structure; it changed how IBM made decisions in ways that created room for the entire PC ecosystem.

Then, in 1982, the Reagan Administration dropped the case as “without merit.” Fifteen years of litigation, no verdict, no remedy. By the standard metrics of antitrust enforcement, the IBM case was a failure.

But the discipline had already worked. The threat of enforcement produced preemptive behavioral change. IBM unbundled software, declined acquisitions, chose non-exclusive contracts, all because its lawyers feared what might happen if it didn’t. The case was dropped, but the ecosystem it created remained.

Wu presents these facts without drawing a strong conclusion about policy design. But the implication is worth noting: the threat of enforcement may matter as much as enforcement itself. The IBM case worked because it was credible, because the telecommunications pattern had established that the government would actually pursue monopolists, because AT&T had been broken up and Standard Oil before that. Maintaining that credibility may require actual enforcement over time, even if individual cases don’t always succeed.

The International Comparison. The United States, Japan, and Britain all started with national champion computing firms: IBM, Fujitsu and NEC, ICL. They made different policy choices.

Japan and Britain protected their champions from domestic competition, hoping to strengthen them against international rivals. The United States subjected IBM to intense antitrust pressure. The results diverged sharply.

Japan and Britain failed to develop significant adjacent industries, particularly software. Their national champions survived but their computing ecosystems did not flourish. The United States was the only country to actively force competition in computing through significant and ongoing antitrust interventions. It was also the only country that developed a dynamic computing industry.

Aerospace: The Counter-Example. Aerospace represents what happens when enforcement is actively discouraged.

In 1997, Boeing proposed to acquire its main domestic rival, McDonnell Douglas. Under prevailing merger standards, the combination was presumptively illegal. But the Clinton administration signaled support for the merger as a means of strengthening Boeing against Airbus. The divided Federal Trade Commission declined to act.

The result has been a single American firm in commercial aviation. Wu argues the national champion approach has not succeeded. Boeing’s stock has grown more slowly than the S&P 500 despite decades of government support. The company has shown dis-efficiencies of scale, including the manufacturing defects that led to the 737 MAX grounding.

Most striking is the absence of industrial succession. Boeing was the dominant firm in the 1920s, and it remains the dominant firm in the 2020s. In computing and telecommunications, multiple paradigm shifts occurred; technological leadership changed hands repeatedly. In aerospace, nothing comparable has happened.

Wu poses the comparison vividly: “A time traveler from the 1960s would be shocked and amazed by the advances in today’s computers and communications technologies. Boarding a passenger airliner, however, would be a familiar experience.”

The Growth Numbers

The differential outcomes can be quantified. A study of economic growth from 1960 to 2007 by Dale W. Jorgenson and colleagues shows some, well a lot, differences across industries. The total economy grew at an annual rate of 3.45 percent. Computing and software, industries subjected to intense antitrust pressure, grew at 20 to 35 percent annually, while telecommunications and IT services expanded at 6 to 8 percent per year. The industries granted de facto immunity from antitrust enforcement fared far worse: automotive grew at just 2.4 percent annually, below the economy-wide average, and aerospace managed only 1.2 percent, roughly one-third the overall rate.

The industries subjected to aggressive antitrust enforcement grew far faster than those granted de facto immunity. Automotive grew below the economy-wide average. Aerospace grew at roughly one-third the overall rate.

On that little note, Wu is careful about causation: “Historical causation is complex, and it certainly cannot be claimed that differential government attention was the single factor that made the difference.” But the pattern is consistent. “If nothing else,” Wu concludes, “there is a clear argument that lack of antitrust attention can yield industrial stagnation.”

CHIPS: A Contemporary Test

The CHIPS and Science Act, enacted in 2022, provides a contemporary test.

The CHIPS Program Office had $39 billion to allocate. It could have followed the national champion model, picking a single firm to back. The Department of Defense’s recent MP Materials agreement took essentially that approach: price floors, demand guarantees, loan financing, and equity investment for a single rare earth producer.

CHIPS took a different path, one that echoes the computing and telecommunications models. Sara O'Rourke, the former COO, describes the process in an article for Factory Settings. The pre-application phase created discipline through iterative feedback: “Another applicant submitted a project for a new technology that seemed to be a minor improvement on an enterprise technology use case, which we indicated would be low priority for us; they returned with a proposal that would represent a total breakthrough in AI datacenter chip technology.”

The process maintained discipline through iterative engagement, preserving the feedback loops that make competition work. In the pre-application phase, CHIPS teams met regularly with applicants to provide feedback and push for more ambitious commitments. Leading-edge applicants initially requested over $70 billion in collective funding. CHIPS got them to accept approximately $26 billion for multiple clusters while extracting commitments to build faster, add more capacity, and push to more cutting-edge technology.

The result was a portfolio, not a champion. CHIPS funded all five companies at the leading edge (TSMC, Samsung, Intel, Micron, and SK Hynix) plus suppliers and advanced packaging facilities. Mike Schmidt, the former program director, contrasts this with the MP Materials approach: CHIPS built “a diversified portfolio,” not “a massive brute-force subsidy to a single player.”

---

Implementation Requirements

Competition does not operate in a vacuum. Effective discipline requires institutional infrastructure.

Institutional Capacity

The Fang study found that high-capacity regions achieved 2.3 times greater productivity effects from identical policies compared to low-capacity regions. Administrative sophistication (the ability to evaluate projects, negotiate with sophisticated companies, structure accountability) determines whether industrial policy succeeds or fails.

The CHIPS program invested heavily in this capacity, hiring what O’Rourke called “top Wall Street and industry talent who understood commercial diligence and deal negotiations.” The telecommunications interventions similarly required sophisticated regulatory capacity at the FCC to design and enforce the Computer Inquiries.

Countries without strong enforcement institutions may not replicate results simply by adopting policies.

Policy Bundles

The Fang study also found that policies employing three or more complementary tools boosted firm productivity 2.3 times more than single-tool approaches.

The successful American technology industries received exactly this: research funding through DARPA and defense contracts, infrastructure investment, education pipelines from universities, all combined with antitrust discipline that maintained competitive pressure. Aid and discipline operated together, not as substitutes.

CHIPS exemplified this bundling. The program deployed pre-application feedback, milestone-based payments, continuous engagement with applicants and their customers, and tailored commitments for each firm. This was aid structured to maintain competitive pressure and accountability.

Maintaining Commercial Ecosystems

On the topic of tech commercialization, we have a cautionary tale in the form of Extreme Ultraviolet (EUV) lithography . EUV, the technology enabling the most advanced semiconductor manufacturing, was largely developed in the United States. DARPA, Bell Labs, the US National Laboratories, IBM Research, and Intel all contributed decades of research and billions of dollars.

Yet today, EUV lithography machines are built by a single firm: ASML, in the Netherlands.

By the time EUV was ready for commercialization, American lithography equipment firms had been almost entirely forced out of the global market. The US did the research but lacked domestic firms positioned to commercialize it.

Antitrust enforcement that preserves competitive entry protects the ecosystem in which commercialization occurs. Research funding alone is not enough if the commercial ecosystem dies. The telecommunications pattern, sustained intervention to keep adjacent markets open, matters precisely because it maintains the firms that can turn research into products.

The Limits

We should be clear about what economists view as genuine constraints.

The Inverted-U’s Downward Slope

Aghion’s model shows that too much competition can reduce innovation. In very fragmented markets, firms may underinvest in R&D because they cannot appropriate returns. The escape-competition effect depends on firms being close enough in capability that pulling ahead seems achievable; when markets are too fragmented, no one can pull ahead far enough to matter.

China’s solar panel industry illustrates the danger. When multiple provinces simultaneously targeted solar with similar subsidies, the resulting overcapacity undermined the industry. Firms proliferated, margins collapsed, and the sector required painful consolidation. China’s 169 automakers may be past the optimal point; current EV production capacity of 54 million vehicles exceeds domestic demand of roughly 30 million.

The feedback that competition provides can itself signal when markets have become too fragmented: collapsing margins, unsustainable losses, inability to fund R&D. The discipline cuts both ways.

Path Dependence

Early choices lock in patterns that persist for decades. The Fang study found that “pioneer” cities, the first to target a sector, achieved 31 percent higher productivity gains than followers.

Path dependence (using the phrase again) cuts both ways. Decisions to maintain competition can produce lasting benefits: the telecommunications interventions that kept AT&T in its lane created conditions that enabled entire industries to develop. But decisions to permit consolidation may be very difficult to reverse. The window for effective intervention may be narrower than policymakers assume.

The non-filing against GM illustrates the point. By the time Japanese competition forced adaptation in the 1970s, decades of limited domestic competition had shaped the industry’s capabilities, culture, and cost structures. The Big Three’s attempts to build competitive small cars failed partly because the organizations had optimized for something else entirely.

Industry-Specific Characteristics

The automobile and aerospace industries have distinctive features (high capital intensity, long product cycles, significant scale economies) that may limit how much competition they can sustain. The relevant question is not whether all industries should be maximally fragmented but whether competitive pressure within each industry’s constraints produces better outcomes than protection.

The evidence suggests it does. Even in capital-intensive automobiles, the fragmented Chinese EV market outperformed the consolidated American industry. Even in scale-dependent aerospace, the absence of domestic competition correlates with stagnation or even decay rather than strength.

The Policy Implication

What follows for policymakers?

First, antitrust enforcement should be understood as one tool in a broader industrial policy toolkit, a form of discipline that differs from subsidies and tariffs but shares the goal of strengthening industry.

Second, a country can simultaneously strengthen an industry through subsidies and research funding while maintaining competitive discipline through antitrust enforcement. The federal government did exactly this in telecommunications and computing. CHIPS did it in semiconductors. The apparent inconsistency is no inconsistency at all. Like an Olympic athlete, an industry can be both supported and challenged.

Third, the failure to enforce is itself a policy choice with consequences. The decision not to file against GM, the decision to permit the Boeing-McDonnell Douglas merger, the broader pattern of treating consolidated industries as national champions: these were choices. They had effects. Non-enforcement should be evaluated with the same rigor as enforcement.

Fourth, credibility may matter as much as cases filed. The IBM experience suggests that the threat of enforcement can produce behavioral change even when cases are never won or are ultimately dropped. But the threat must be credible, which requires a history of actual enforcement. The telecommunications pattern, seventy years of sustained pressure, created expectations that shaped corporate behavior even in the absence of active litigation. Wu doesn’t draw strong conclusions about how to design for this effect, but the implication is that institutional commitment across administrations matters.

Fifth, (and you know I love my dead horses) implementation matters as much as policy design. Competition requires institutional capacity, policy bundles that combine complementary tools, and sustained commitment over decades. The telecommunications pattern produced results that a single intervention could not have achieved.

Sixth, maintaining commercial ecosystems matters as much as funding research. Antitrust enforcement that preserves competitive entry protects the firms that can turn innovations into products. Without that ecosystem, research investments may benefit other nations’ industries.

Back to the Beginning

We return to Honda, and to the provincial officials in Guangdong who never meant to prove him right.

Honda’s bet was that competitive pressure would produce better outcomes than centralized coordination. He wasn’t making an ideological argument about markets versus government. He was making a claim about feedback: that the discipline of competition, the constant signal that you must improve or lose, would drive innovation more reliably than protected consolidation. The Chinese provincial officials who blocked Beijing’s merger plans weren’t making any argument at all; they were protecting their tax bases and their careers. But the effect was the same. By preventing consolidation, they preserved the competitive pressure that forced Chinese automakers to innovate.

The subsequent sixty years have rendered a verdict. The fragmented Japanese auto industry, which Honda helped preserve, produced globally competitive manufacturers. The fragmented Chinese EV industry, which provincial resistance preserved despite Beijing’s plans, has produced the world’s leading electric vehicle companies. The American auto industry, which achieved the consolidation that MITI sought for Japan and Beijing sought for China, required government bailouts.

In computing and telecommunications, sustained antitrust pressure produced industrial succession, the renewal that Schumpeter identified as central to capitalist dynamism. The leading firms of 2025 are unrecognizable from those of 1965. In aerospace and automotive, where antitrust attention faded, the same firms that dominated in the 1950s dominate today.

Within governments, separate teams typically work on “competition” and “competitiveness.” The evidence suggests these should be the same conversation. Competition creates competitiveness. Forcing an industry to compete internally, maintaining the feedback loops that discipline performance, makes it more competitive in the world.

We do not claim fragmentation is always preferable to consolidation, or that antitrust enforcement always promotes innovation. Never been a fan of the “one weird trick” theory of anything. At the end of the day, the relationships are contingent and context-dependent. But the evidence is sufficient to reject the conventional wisdom that antitrust and industrial policy are opposing strategies.

Competition is not the enemy of industrial strength. It is, properly understood, itself a form of industrial policy.