By 1990, the world economy was deep into a new phase of globalization: trade barriers were falling, multinational supply chains were thickening, and capital moved more freely than at any time since the pre-1914 era. Yet the dominant analytic tool for long-run growth—the neoclassical model associated with Robert Solow’s 1956 paper “A Contribution to the Theory of Economic Growth”—left an uncomfortable gap. In Solow’s framework, sustained per-capita growth ultimately comes from “technical change,” but the model treats that driver as exogenous: it arrives from outside the economic system. For policymakers confronting widening international income gaps, the rise of East Asian industrialization, and debates over education and innovation policy, “technology” looked less like manna and more like something produced—unevenly—by firms, universities, and states. Endogenous growth theory emerged to address precisely that puzzle: how can market incentives, investment decisions, and institutions generate ongoing technological progress and persistently different growth rates?
Two figures were central to this shift: Robert E. Lucas Jr. and Paul M. Romer. Lucas’s 1988 paper, “On the Mechanics of Economic Development,” placed human capital accumulation at the center of development and growth, emphasizing learning and spillovers. Romer’s 1990 paper, “Endogenous Technological Change,” gave a formal, micro-founded account of research and development (R&D) and “ideas” as inputs—nonrival goods that can generate increasing returns. In different ways, both responded to what Lucas called the “central question of economic growth”: why do some countries grow rich while others remain poor, even under globalization’s expanding opportunities?
At the core of endogenous growth is a reframing of what an economy accumulates. In the neoclassical tradition, the key state variables are physical capital and labor, operating under diminishing returns. Endogenous growth adds forms of capital that behave differently: human capital (skills, knowledge embodied in workers) and ideas (designs, blueprints, scientific principles, algorithms)—knowledge that can often be used by many at once. This distinction matters because nonrivalry changes the arithmetic of growth. A machine can only be used in one place at one time, but a chemical formula or a software protocol can be replicated at near-zero marginal cost across the globe. Under globalization, that replicability is amplified: once an idea is created, it can be diffused—imperfectly—through trade, foreign direct investment, migration, and communication networks.
Lucas (1988) formalized an economy where individuals invest in education and learning, increasing their own productivity while also raising the productivity of others through spillovers—what later economists would call “externalities.” The key is that the social return to accumulating human capital can exceed the private return, leading competitive markets to underinvest. Romer (1990) pushed further into the production of ideas themselves. He separated the economy into a final-goods sector and an R&D sector that produces new “designs” used to create differentiated intermediate inputs. Because each new design expands the set of available inputs, the productivity of the final sector rises with the stock of knowledge. Romer emphasized two linked propositions: (1) ideas are nonrival, enabling increasing returns at the aggregate level, and (2) because ideas are partially excludable via patents or secrecy, private firms can be incentivized to produce them, but will still tend to produce too few relative to the social optimum.
The mathematical backbone of these models is intentionally spare, designed to show how constant (or rising) returns to accumulable factors can sustain long-run growth without exogenous “technology shocks.” A canonical Lucas-style setup writes output as $$ Y(t)=K(t)^{\alpha}\big(u(t)H(t)\big)^{1-\alpha}, $$ where (K) is physical capital, (H) is human capital, and (u\in(0,1)) is the fraction of human capital devoted to production (with (1-u) devoted to education). Human capital evolves as $$ \dot H(t)=\phi\,(1-u(t))\,H(t), $$ implying that the growth rate of (H) is proportional to the share of time spent learning. Because (H) enters production, sustained investment in learning can sustain growth in (Y) even if physical capital faces diminishing returns. Many variants add an externality term—e.g., average human capital (\bar H)—to capture spillovers, making the aggregate production function effectively constant returns in reproducible factors even when private firms perceive diminishing returns.
Romer’s 1990 model makes the stock of ideas explicit. A simplified representation writes final output as $$ Y(t)=L_Y(t)^{1-\alpha}\left(\int_{0}^{A(t)} x(i,t)^{\alpha}\,di\right), $$ where (A(t)) is the number of available designs (a measure of knowledge), (x(i,t)) is the quantity of intermediate input (i), and (L_Y) is labor in final production. New ideas are produced in R&D according to $$ \dot A(t)=\delta\,A(t)^{\phi}\,L_A(t), $$ with (L_A) labor allocated to research. The term (A(t)^{\phi}) captures the “standing on shoulders” effect: existing knowledge can make it easier to produce new knowledge. If (\phi=1), knowledge growth is proportional to existing knowledge and research effort; if (\phi<1), diminishing returns to research can set in; if (\phi>1), growth can become explosively fast (usually viewed as unrealistic without countervailing forces). The key insight is that because (A) is nonrival, it can enter production without being used up, and the expansion of varieties (A(t)) can offset diminishing returns to physical inputs, generating sustained per-capita growth.
Historically, endogenous growth theory reshaped research agendas and policy debates in the 1990s, a decade marked by the WTO’s creation (1995), the TRIPS agreement on intellectual property, and the rapid spread of information technologies. Growth accounting—once a largely descriptive exercise decomposing output growth into factor accumulation and a residual—was complemented by models tying the residual to measurable activities: schooling, R&D spending, patents, and innovation systems. It also changed how economists interpreted globalization. Instead of viewing openness mainly as a way to allocate resources more efficiently at a point in time, endogenous growth models emphasized dynamic gains: trade and investment can speed the diffusion of ideas and raise the incentives to innovate by enlarging markets.
Criticism came from several directions. Empirically, early cross-country growth regressions that appeared to validate strong policy levers were later questioned for fragility and identification problems. Theoretically, some objected that early models could generate “scale effects”—the prediction that larger economies should grow faster because they have more researchers—whereas in the data, growth rates do not rise mechanically with population. Later work (including semi-endogenous growth models) modified the idea-production function to reduce or eliminate these scale implications. Others argued that patents and market power, central to Romer’s mechanism of appropriating returns to ideas, might create static inefficiencies even as they encourage dynamic innovation—posing a difficult tradeoff for policy.
The legacy of Romer and Lucas is evident in today’s economics of innovation, inequality, and industrial policy. Modern models of “directed technical change,” innovation networks, and endogenous automation extend their logic: incentives shape which ideas are produced and who benefits. The policy implications remain highly contemporary: education and training, basic research funding, immigration of skilled workers, competition policy in technology markets, and intellectual property design all affect the economy’s idea-production engine. As global supply chains fragment and data and AI become strategic assets, the notion that “ideas are inputs” has only gained salience.
In the longer line of continuity, endogenous growth sits between the postwar neoclassical synthesis and later micro-empirical innovation economics. It inherits from Solow the discipline of aggregate modeling and balanced growth, but rejects the assumption that technology is exogenous. It also draws from Joseph Schumpeter’s earlier emphasis on innovation and creative destruction, translating that vision into general equilibrium models with explicit incentives and spillovers. After Romer and Lucas, the frontier moved toward richer firm dynamics and measurement—linking patents, productivity, and market structure—and toward the institutional foundations of innovation. In that sense, the 1990 moment was pivotal: amid globalization’s acceleration, growth theory re-centered on the economics of ideas, making technological change not a residual to be assumed, but a process to be explained.