Revisiting the Determinants of Economic Growth Theory

Endogenous growth models delivered a powerful advance by putting ideas and innovation at the center of long-run prosperity, earning renewed attention over the past year as the focus of the Nobel Prize in Economics. These models emphasize sectoral reallocation, R&D activi­ties, human capital, and nonrival ideas as the central engines of sustained growth, often overlooking the institutional and social conditions under which innovation is produced, diffused, and translated into broad economic gains.¹ In doing so, they often treat the surrounding social environment as largely fixed, or at least slow-moving relative to technological progress itself.

While these mechanisms undoubtedly play an important role, and the theoretical contributions of these studies have shaped the economic landscape, economic growth today is not well described by a frictionless pipeline from discovery to diffusion—it is challenged by an array of frictions, ranging from brittle production networks to atrophied capa­bilities to weakened social and cultural capital across nations. Even though concerns about supply-chain resilience became elevated during the Covid-19 pandemic, they have been known for over two decades. A 2013 study led by MIT’s Production in the Innovation Economy found “gaping holes” in the U.S. production network.² That was also the year that the seminal “China Shock” paper was published by David Autor, David Dorn, and Gordon Hanson, explaining how export competition with China led to a deterioration of the middle class in the United States.³

In a nutshell, this is why the industrial policy debate matters for growth, not as a detour into protectionism or a search for “critical sectors,” but as a question about how to rebuild the nodes and linkages that make people productive and promote human flourishing. The endogenous growth models that won the Nobel Prize are worthy of celebration, but they are inadequate at explaining the world of 2026 and beyond; this is because the array of simplifying assumptions such models were based on, however historically reasonable they may have been initially, are now harder to swallow.

For instance, labor supply, family formation, and demographic reproduction typically enter as background conditions rather than objects of analysis, so changes in fertility or labor force participation are treated as exogenous to the growth mechanism. Human capital is usually modeled as accumulating through investment and incentives, which keeps the theory tractable but tends to abstract from the social context in which education and skill formation actually occur. In the modern U.S. economy, these factors have become more variable across places and groups, and they matter for whether ideas translate into broad-based productivity gains.

Economic historians pointed out years ago that the unexplained residual in growth accounting likely hides the influence of unmeasured social factors. And while endogenous growth models have made plenty of advances, they still assume the broader social environment is constant and generally take the production network as fixed. Sociologists and political scientists have long quantified the role of social capital (trust, norms, networks) in economic performance, but these are often bolted on as afterthoughts rather than integral parts of growth models. We in economics often describe these omissions as “missing state variables,” key conditions that are not captured in the formal dynamics of the model, and instead explain that our “stylized model” is designed to capture a narrow set of patterns in the data. But the fruit of our efforts is a theoretical apparatus that looks under the lamppost of R&D and physical capital, while much of what determines how ideas are translated into output remains in the shadows.

A Systems Approach to Economic Growth

It is a mistake to define industrial policy narrowly as writing checks to manufacturers. In mature economies, the most consequential industrial policy often operates through the institutional plumbing that determines whether investment is feasible at all. That includes workforce development systems that match training to real production needs, infrastructure build-out and maintenance, regulatory processes that can approve and monitor projects without endless delay, public procurement that creates predictable demand for new capabilities, and regional strategies that anchor supplier ecosystems rather than scattering them. Industrial policy in today’s economy is less about choosing winning firms and more about rebuilding the state and market interfaces that allow private initiative to scale. It targets the points where the economy repeatedly fails to convert promising innovations into widespread productivity improvements.

But we should think about institutional capacity broadly as the scaffolding. The same growth process depends on social inputs that shape labor supply, skill formation, and community-level resilience. If social and relational factors are part of the hidden infrastructure of growth, family stability is a central component. A growing body of evidence links shifts in family structure, including rises in single-parent households, declining marriage rates, and lower fertility, to measurable changes in labor market attachment, human capital accumulation, and community outcomes that ultimately feed into economic performance. These are not moral claims. They are mechanisms through which family structure shapes the productive capacity of society.

Consider labor force participation, especially among prime-age men. Their labor attachment in the United States has been declining for decades, and delayed or foregone marriage appears to be an important piece of this puzzle. Married men have significantly higher labor participation and work hours than unmarried men, even after controlling for education and other factors. Part of this is selection, but part of it is also causation: marriage increases a man’s economic motivation by creating an incentive to work harder and more steadily to support a family.

In particular, a recent study models these phenomena structurally by simulating the labor market with late-1990s marriage rates versus today’s; it found that a 2.3 percent reduction in average hours worked by prime-age men is attributable to the marriage decline.⁴ When fewer men shoulder family responsibilities, a nontrivial number opt out of full‑time work. Stable family bonds have long been known to bolster labor force attachment, providing structure and support against idiosyncratic shocks.⁵ Conversely, when family formation falters, we see more men on the margins of the job market.⁶ This linkage remains largely invisible in macro models that take labor supply as given, but it is palpable in communities across the country.

Family stability (or the lack thereof) also feeds directly into fertility and long-run demographics, literally shaping the quantity of future labor and consumers. The United States, like many advanced economies, has seen a steep decline in birth rates, now well below replacement level. Popular discourse often attributes this to young people’s changing preferences or economic anxieties, but a more concrete driver is the decline of marriage. Marital status remains strongly predictive of childbearing: the birth rate for married women is more than twice that of unmarried women, and completed fertility is far lower among women who never marry.⁷ Demographers have shown that “essentially all of the decline in U.S. fertility since the 2000s can be explained by changes in the marital composition of society.”⁸ If the age-specific marriage rates of, say, 2000 had prevailed in the 2010s, American fertility might have been substantially higher.

In practical terms, stable marriages facilitate childbearing—through combined incomes, sharing of child-rearing, and greater optimism about the future—whereas unmarried individuals are increasingly foregoing parenthood. Industrialized countries are learning that once fertility expectations reset lower, it is hard to turn that ship around. Thus, weakening family formation is not just a cultural trend. Current growth theory—barring some work by Chad Jones—treats fertility as exogenous and immaterial to growth. Yet if half of young women never have a child, that is a fundamental economic shift as well. A nation’s innovative capacity means little if there are too few hands to build and minds to implement those innovations a generation from now.

Family structure’s impact on human capital formation is equally as direct. A stable family provides not only resources for children, but also the socialization and emotional support that foster both cognitive and “soft” skills. Decades of research in sociology and economics find that children raised in single-parent or highly unstable family settings have worse educational and human capital outcomes—lower test scores, higher dropout rates, and fewer years of schooling—which has been validated using recent and modern quasi-experimental techniques.⁹ They are also more likely to exhibit noncognitive and soft skill deficits that can hinder success in adulthood.

These gaps persist even after controlling for income and other variables, suggesting that family stability itself confers advantages through greater parental attention, consistency in routines, and access to social capital. The rise in single parenthood since the 1970s has thus been linked to a slowdown in human capital accumulation at the aggregate level, even if test scores overall have improved due to other factors. Moreover, family breakdown often creates a cycle: lower parental resources and time investment impede a child’s educational attainment, which in turn affects that child’s economic prospects and family stability in adulthood. This dynamic feedback loop is hard to capture in growth equations, but it undoubtedly plays out in the real world, dragging down productivity growth over time. Importantly, these effects are not uniform—many single parents heroically raise thriving children—but at scale, the probabilities shift in a direction that hampers the development of a highly skilled, adaptable workforce. As Nobel laureate James Heckman has emphasized, skill formation is a cumulative process starting in early childhood; disruptions in the first environment of learning, the family, can have lifelong economic consequences.¹⁰

Beyond individual outcomes, family stability (or instability) influences community resilience and social capital, which also impact productivity. Communities with a higher share of intact families tend to have stronger neighborhood networks, higher trust, and more civic engagement, classic ingredients of social capital that make coordinated economic activity possible. Leading empirical work by Raj Chetty and his Opportunity Insights Lab has found that the fraction of children living with single parents is the single strongest correlate of low upward income mobility in a community.¹¹ In other words, areas with more fragmented families see poorer long-run outcomes for children across the community because social ties fray and collective efficacy diminishes. In fact, the prevalence of single-parent households in a county is a better predictor of a child’s chances of escaping poverty than factors like school quality or income inequality. High-marriage communities often marshal informal support systems—neighbors helping with childcare, coaching Little League, enforcing norms—which can “weather the storm” during hard times, including most recently over the Covid-19 pandemic.

In a recent study of mine and Clara Piano’s using data from the Gallup World Poll from 2006 to 2025, we find that when people are more optimistic about their future, they are more likely to have children.¹² This relationship holds even when incomes, education, employment, and marriage rates are taken into account. Put simply, people are more willing to start or expand families when the future feels predictable and worth investing in, and they pull back when it does not. Fertility is not only a response to wages or subsidies; it reflects whether households believe their society is on a stable and upward path.

Growth today is constrained not only by the pace of innovation, but by whether economic systems generate the confidence, stability, and institutional support that allow people to make long-term commitments. Industrial policy, understood narrowly as subsidizing production, cannot address this problem. But industrial policy understood as rebuilding the economic and social foundations that make the future feel viable can. When policies restore reliable employment pathways, reduce volatility, and rebuild the connective tissue of local economies, they do more than raise productivity in the short run. They help reestablish the conditions under which families form, human capital accumulates, and growth becomes self-sustaining again.

This begs the question: why do such obviously crucial factors remain invisible in cross-country growth regressions and macro models? Part of the reason is measurement difficulty and modeling conventions. Family stability and social cohesion are hard-to-quantify qualities, and economists have often lacked consistent cross-country data on them (i.e., trust surveys and marriage statistics exist, but they are tricky to compare and incorporate). Many growth studies include proxies, like years of schooling or rule-of-law indices, but not variables for “community integrity” or “two-parent household share.” The elements that we can most readily measure (years of education, R&D spending, physical investment, etc.) end up as the favored regressors, whereas softer social indicators either get left out or show up only in qualitative discussions.

Furthermore, the effects of family and social variables tend to play out over long horizons and interact with cultural context, making them harder to disentangle causally. A country with historically strong families and one with weaker families may also differ in myriad other ways (religion, policies, history, etc.), so simple cross-country comparisons can be confounded. Traditional models also assume optimizing behavior in a way that abstracts from social constraints. For example, a representative household model might assume everyone who should work will work, and everyone who should invest in education will do so, barring policy distortions. Such models have a tough time explaining phenomena like prime-age men dropping out due to despair or lack of familial support, or teenagers not pursuing college because of absent parental guidance. Those are considered outside the economic model: exogenous preferences or idiosyncratic shocks.

In cross-country regressions, if one tries to include a variable like “fraction of children in single-parent homes,” it might be highly collinear with other development indicators, or simply deemed a cultural trait rather than a policy-handle variable, and thus omitted. The upshot is that the real constraints imposed by family instability and social decay are “felt but not seen” in mainstream growth analysis. They show up as lower productivity or fewer workers than expected, without an explicit causal tag. This misdiagnosis can lead policymakers to focus on symptoms, rather than the root cause (i.e., they might consider giving subsidies for innovation or jobs training when the binding constraint might be social).

Economic Acceleration through Concentration

Economic history shows that concentrated, tangible investments—in physical infrastructure, industrial capacity, and strategic technology—have repeatedly driven surges in growth. Many of our greatest growth accelerations were not just products of decentralized market forces, but of purposeful, large-scale projects that unlocked productivity.

A classic example is America’s nineteenth-century push to build out canals and railroads. The Erie Canal completed in 1825 is often cited as one of the most impactful infrastructure projects in U.S. history: it opened the Midwest’s abundance to East Coast markets, slashing transport costs and spurring settlement and development across large swaths of the country. By connecting previously isolated regions, the canal did not just help the “transportation sector”; it boosted agriculture, commerce, banking, and even industrial development along its route.¹³

This builds on the modern empirical literature on transportation infrastructure, including work by Dave Donaldson and Richard Hornbeck, showing that railroads raised local market access and generated large, economy-wide gains by integrating regions into national and international markets.¹⁴ For example, the Pacific Railway Act of 1862 provided federal land grants and loans to build the first transcontinental railroad, knitting together the coasts during the Civil War. The railroad was a textbook case of a network with massive spillovers: once the rail links existed, farms, mines, towns, and factories flourished along them, creating value far beyond the revenues of the rail companies themselves. These investments were strategic concentrations of capital under conditions of national stress, expanding the production frontier and bringing new resources and people into the economic sphere, enhancing productivity through scale and specialization. Such endeavors often required overcoming a coordination problem or an initial loss that private actors were loath to bear, hence the critical role of public policy in spurring them.

The deeper point is that America’s later manufacturing dominance was not built by private initiative alone. Long before the twentieth century, the federal armories at Springfield and Harpers Ferry served as laboratories for standardized production.¹⁵ Through sustained procurement, in-house engineering, and close interaction with private contractors, the armories advanced the use of interchangeable parts, precision gauging, and specialized machine tools in firearms production: developments often grouped under the “American system of manufacturing.” Over the nineteenth century, these techniques and the skilled mechanics who mastered them diffused beyond armaments into civilian and consumer industries.

From the vantage point of industry, how the United States handled World War II production was far from perfect; however, there were several outstanding instances of manufacturing efficiency. For example, the War Production Board coordinated materials allocation and production priorities at a national scale and helped to mobilize a rapid expansion of industrial capacity. That mobilization also created durable capabilities. The U.S. synthetic rubber program, for example, combined federal coordination with private production and scientific expertise to build an industry essentially from scratch within a few years.¹⁶ Recent research has also found that wartime R&D through the Office of Scientific Research and Development (OSRD) led to the formation of technology clusters across the country and increases in high-tech entre­preneurship and employment.¹⁷

In the postwar era, the Interstate Highway System was a similarly instructive case. The 1956 Federal Aid Highway Act authorized the laying of forty-one thousand miles of road for a national interstate network, effectively creating a new backbone for freight and passenger movement. Economically, the point is not simply that highways made travel easier. They changed the feasible set for logistics, inventory management, and regional specialization. During the peak buildout years, productivity growth rose disproportionately in industries that relied more heavily on vehicles, a pattern that supports a causal channel from roads to productivity rather than the reverse. As the network approached completion, the incremental boost naturally diminished, which is a reminder that these large physical platforms often raise the level of productivity in discrete steps.

The key features of these successes are network effects, choke point alleviation, and multiplier effects. By network effects, I mean the self-reinforcing value that grows as a network is completed. Every new rail spur or internet node enables new connections and economic activities in a nonlinear way. Unlike diffuse, small-scale spending, these kinds of concentrated investments often create clusters of economic activity. These clusters yield knowledge spillovers and attract complementary businesses. They are far more than the sum of their parts.

This is also one of the reasons that current investments in AI are not going to show up fully in productivity statistics; the fruition of their full value comes after complementary investments have been made. The best way to see the effects of technology investments in the short run is to look at the composition and performance of the labor market in the associated industries.

Accordingly, darpa exemplifies the benefits of this modern industrial policy. Many frontier technologies look unattractive to private investors at the outset, not because they lack promise, but because the early problems exist at the system level: interoperability, standards, complementary infrastructure, and a long chain from prototype to usable capability. Government can act as the first mover that coordinates dispersed actors, sets technical objectives, and provides patient early demand, which can then create an ecosystem that attracts subsequent private investment. That logic is exactly what led the Advanced Research Projects Agency (ARPA) to produce breakthrough technologies that would have not otherwise been achievable through private capital alone, most notably the internet. The precommercial network emerged under darpa sponsorship well before market incentives could plausibly support it, at a time when the technology lacked clear business models and required coordination across universities, contractors, and carriers.¹⁸

Two current U.S. initiatives illustrate how the state is already acting as a first mover for AI and the energy infrastructure that will sustain it. The Trump Administration’s AI Action Plan explains that winning the AI race is about building “vast AI infrastructure and the energy to power it,” treating data centers, semiconductors, and energy supply as strategic complements rather than separate policy domains.¹⁹ In addition, the Genesis Mission launched as a coordinated national effort led by the Department of Energy to build an integrated platform that links federal scientific datasets, high-performance computing, and AI systems, with an explicit mandate that includes “secure energy dominance” and the use of existing research infrastructure and production assets. These policies reflect targeted collective action aimed at relieving binding constraints, coordinating investment across a fragmented landscape, and establishing the institutional and physical platforms that private actors can build on. That is increasingly what economic and technological sovereignty looks like in practice: mobilizing infrastructure, data, standards, and capabilities at scale when markets alone cannot coordinate the leap.

These examples, and many more, challenge the reflexive “Smoot-Hawley bad” narrative that any industrial policy or strategic intervention must inevitably lead to protectionist disaster. The historical record of U.S. growth is not one of laissez-faire purity occasionally interrupted by foolish protectionism. Rather, it features recurring episodes of active public guidance and investment to build new capacity, often hand-in-hand with private enterprise. Importantly, the successful episodes were usually those that created new value rather than those that merely shielded existing industries behind tariff walls. Industrial policy, when it takes the form of supply-side investment and strategic support can expand the economic pie rather than just reallocating slices.

To be sure, the United States has had its share of policy misfires and rent-seeking under the guise of industrial strategy. But the notion that any deviation from pure free market policy equals “picking winners” and doomed favoritism is not borne out by the American experience. On the contrary, some of the most fertile periods of American innovation and growth were catalyzed by deliberate government action. History suggests that strategic concentration can be a form of economic dynamism, not stagnation, especially in the midst of geopolitical competition.

Industrial Policy Reframed

In the traditional view, industrial policy is justified either by economics (e.g., fixing R&D spillovers or capital market failures) or by national security (e.g., ensuring domestic capacity in defense-critical areas). I propose an expanded framing: industrial policy as a mechanism to stabilize labor markets, rebuild skills, and foster family and community formation in an era where those relational structures have become the true bottlenecks to growth. In other words, the aim is not just more silicon chips or solar panels per se, but the revitalization of the human and physical ecosystems that translate innovation into widespread productivity.

This reframing shifts the focus from abstract output metrics to the lived economic experience of American communities. A policy that supports, say, the establishment of a manufacturing hub in a struggling Midwest town can be seen through this lens: it’s not only about the output of that factory, but about providing steady employment for households, restoring a tax base for public schools, and giving young people a reason to settle down and raise families. Such an industrial policy has an added benefit of serving as place-based social policy. By anchoring good jobs in areas hollowed out by deindustrialization, it addresses the social distress (opioid abuse, family breakdown, out-migration, etc.) that has held back U.S. productive potential. It’s a strategy of healing after injury, treating communities that lost their economic lifeblood not with one-time aid, but with new arteries of industry that can sustain life for the long term.

To make this concrete, consider how the “systems” framing changes what counts as a credible rationale for public support. Instead of saying, “We must subsidize advanced manufacturing because of spillovers in innovation,” the argument becomes, “We invest in specific capabilities and places because they rebuild the pipelines that turn technological possibility into broad-based participation.”

That platform logic is central to understanding why advanced manufacturing rarely returns on its own. Modern manufacturing competitiveness is no longer driven primarily by low unit labor costs, but by the ability to design, simulate, test, certify, and continuously improve complex production processes. Those capabilities depend on shared digital infrastructure, access to high-end computing, standardized data, quality systems, and specialized testing facilities that no single firm has the incentive or capacity to build in isolation. In their absence, production gravitates toward countries that compete through regulatory arbitrage, and exploiting weaker labor, environmental, or safety standards rather than superior capability.

An initiative like Genesis Mission, as discussed earlier, is a useful contemporary example since it targets this coordination failure. By lowering the fixed costs of digital design, process optimization, and validation, it enables firms and regions to compete on precision, reliability, and speed rather than on labor exploitation. This is also the mechanism documented in my work with Giovanni Gallipoli, which shows that the expansion of digital tasks within manufacturing is what underpinned its productivity resurgence, not the hollowing out of production itself.²⁰ Advanced manufacturing, in this sense, is not a sector that can be subsidized back into existence. It is an ecosystem that has to be rebuilt through shared capabilities before private investment can scale.

Crucially, this reframed industrial policy strengthens the conditions under which innovation and diffusion occur. A workforce that is secure and rooted is more likely to invest in training and adapt to new processes, while regions trapped in decline experience outmigration, firm exit, and the erosion of the civic and relational infrastructure that supports work. Of course, we will need measurement strategies for evaluating the efficacy of these interventions, and that is exactly what the Global Flourishing Study—a joint initiative between Gallup and researchers at Baylor and Harvard universities—is doing by surveying individuals longitudinally across the United States (and the world) along a multidimensional set of individual-level mental, physical, and social outcomes.²¹

Transparency and Governance

If industrial policy is to take on an ambitious role in rebuilding economic and social capacity, it must be designed around transparency, conditionality, and accountability. A persistent failure is that programs become open-ended entitlements or are captured by special interests. Avoiding those traps requires governance mechanisms from the start: clear objectives, metrics tied to those objectives, and explicit exit rules when goals are not met.

Historical precedents and modern proposals point to the same discipline: support should be contingent on performance. In South Korea’s export-oriented industrialization, state support was tied to concrete export targets. Firms that failed to meet benchmarks could lose preferential access to credit and other forms of support. This logic of reciprocal obligation matters because it frames public assistance as a contract, not a transfer. Firms, industries, and regions receive support only if they deliver the agreed outcomes, whether those are local investment, training commitments, export performance, or job creation.

Conditionality only works if it is measurable and routinely enforced. Programs should be monitored at least annually, with support continuing when milestones are met and tapering when they are not. This is also the core of Dani Rodrik’s point that the real test of industrial policy is not whether governments can reliably pick winners, but whether they have the capacity to let losers go.²² That requires political and administrative willingness to terminate, redesign, or consolidate interventions that are not producing results.

Sunset clauses and periodic review should therefore be standard. Subsidies, tax credits, tariffs, and procurement preferences should have an explicit expiration date, along with renewal criteria linked to program objectives. An apprenticeship subsidy, for example, might run for five years and be renewed only if participation and skilled trades employment rise in a way that is consistent with the program’s stated goals. Sunsets also make discipline easier in practice because letting a program expire is often less politically costly than repealing it.

A serious governance framework also forces explicit tradeoffs into the open. Supporting domestic production may raise some consumer prices, but that cost should be weighed against the benefits of stronger labor market attachment, skill formation, and regional resilience. Likewise, place-based strategies may shift marginal investment away from already successful hubs. Transparent policymaking should state these tradeoffs plainly and quantify them when feasible, rather than hiding behind either purely technocratic language or purely political appeals.

Finally, a governance framework should treat midcourse correction as normal. Industrial policy should be treated as iterative: evaluate results, scale what works, and revise what does not. If an apprenticeship subsidy fails to attract small firms, the remedy may be to simplify administration or adjust eligibility. If a tax credit is being captured by firms that would have invested anyway, eligibility should be tightened or the credit phased out. The objective is not to avoid failure entirely, but to ensure that programs learn quickly and do not calcify into permanent transfers.

This article originally appeared in American Affairs Volume X, Number 1 (Spring 2026): 143–55.

---

Notes