Scale Effects in Growth Models: What They Are and Why They Matter

A core puzzle in growth theory hinges on a simple observation: richer, larger economies don’t reliably grow faster than smaller, poorer ones. Yet basic endogenous growth models predict they should—larger populations and capital stocks should generate more innovation and faster gains in productivity. This gap between theory and reality motivated decades of refinement. Understanding scale effects in growth models is key to making sense of why development remains unevenly distributed and why policy prescriptions that work at one scale don’t automatically scale up.

The basic scale-effect premise

In standard endogenous growth models (exemplified by the Romer model, circa 1990), larger economies ought to grow faster. The logic runs as follows: growth originates in innovation and knowledge. The more researchers, engineers, and capital goods available—that is, the bigger the economy—the more innovation occurs per year. And if innovation is the engine of economic growth, a larger stock of innovative inputs ought to produce faster productivity gains.

Mathematically, this translates to a prediction: the growth rate of GDP per capita should be increasing in the size of the population or total capital stock. Doubling a country’s research workforce should double its flow of new ideas. Double the total economy and you double growth—a strong scale effect.

This has obvious appeal as a model prediction. It also makes intuitive sense for many real-world domains: a tech hub with 10 million engineers generates more patents than one with 100,000. Yet when economists look at actual growth rates across countries—comparing the U.S. (population 330 million) to New Zealand (5 million), or China today to China in 1980—they find no reliable relationship between size and growth speed.

The empirical puzzle: why large economies don’t grow proportionally faster

If scale effects were strong, we would expect:

1. Larger economies to persistently grow faster. The U.S. economy in 2025 is vastly larger than it was in 1950, yet real per-capita GDP grows at roughly 2–3% per year in both eras—not faster now.

2. Population growth to predict accelerating growth. The world population has roughly tripled since 1950; total global output is vastly higher. Yet per-capita growth hasn’t systematically risen; if anything, it has slowed in many regions.

3. Smaller countries to lag systematically. Luxembourg should grow slower than Germany; Taiwan should grow slower than China by population. In practice, growth rates depend on institutions, openness, and investment—not mainly size.

The empirical record is not perfectly flat (some growth variance does correlate with initial wealth, policy, and trade openness), but the correlation with scale is weak to absent. This disconnect forced theorists to rethink how growth is generated.

Responses: Semi-endogenous and Schumpeterian models

Semi-endogenous growth models (Jones, 1995; Segerstrom, 1998) repair the scale-effect problem by decoupling innovation output from the size of the research sector. In these models, the flow of new ideas is proportional to research effort relative to the existing stock of knowledge. As knowledge accumulates, each researcher must work harder to find an idea—diminishing returns in idea production. The upshot: doubling the research workforce does not double the innovation rate; it increases it, but less than proportionally. And when population grows, the per-capita research effort may actually fall (the innovation pie is spread over more people), leaving growth rates roughly stable.

In semi-endogenous models, the economy still grows in the long run (new ideas continue to arrive), but the growth rate is independent of population size. Larger economies have higher absolute output, but not faster per-capita growth. This aligns with data: China’s per-capita growth has remained in the 5–10% range across different eras and sizes; the U.S. growth rate hasn’t fallen as it has become richer and larger.

Schumpeterian models (Aghion and Howitt, 1992 onward) take a different route: they emphasize competitive dynamics and incentives. In these models, firms race to improve existing products or invent new ones. A firm that successfully innovates captures temporary monopoly rents, driving its investment and research effort. As it innovates again, it pushes out rivals; older product lines become obsolete. The model generates endogenous growth, but the growth rate depends not on how many researchers exist, but on how attractive the innovation prizes are—which depends on market structure, patent length, and the gap between the productivity of the best and second-best firm.

Crucially, the Schumpeterian framework predicts that growth rates are largely insensitive to population size. A huge population of low-skill workers, or a small population of high-skill ones, can generate similar growth rates if incentives are aligned. What matters is the competitive intensity, the ease of entry, and the durability of innovators’ advantages—not scale.

Implications for policy and development

If scale effects are absent or weak—as modern theory and evidence suggest—then the policy levers for growth operate through channels other than mere size:

Institutional quality matters intensely. A well-governed small economy (Singapore, New Zealand) can grow as fast as a large, poorly governed one (Nigeria, Argentina relative to its potential). A country cannot grow faster simply by expanding; it must improve the returns to investment, research, and business formation.

Openness and trade allow smaller economies to tap global markets and knowledge. Taiwan’s growth doesn’t depend on having 1 billion people; it depends on integrating into global supply chains and attracting multinational R&D.

Innovation incentives—patent policy, R&D tax policy, labor mobility—affect growth independent of size. A small country offering strong IP protection and low corporate taxes can attract disproportionate research activity.

Diminishing returns to capital imply that simply adding more labor or capital to a fixed technology cannot sustain acceleration. Growth eventually slows unless productivity improves—and that requires the institutions and incentives to generate new knowledge.

Why this matters for growth debates

The removal of scale effects from modern growth theory has subtle but profound implications. In the 1990s, some observers hypothesized that the digital revolution would permanently raise growth rates in large, capital-rich economies—more workers, more capital, more IT tools should generate faster growth. Yet per-capita growth in the U.S., Europe, and Japan has remained in the 1–3% range despite massive expansion of tech and capital. Scale effects models would have predicted an explosion; institutional and Schumpeterian models predict roughly stable rates unless the fundamental incentives change (which they didn’t in a major way).

Conversely, when a country (South Korea, Vietnam) rapidly raises institutional quality and integrates into global trade, growth can accelerate even with the same or smaller population. This is consistent with incentive-based models and inconsistent with pure scale-effect predictions.

A concrete comparison

Consider two scenarios: Scenario A (1980): China has 1 billion people, a closed economy, weak IP protection, and low capital stock. Growth per capita is roughly 2–3%. Scenario B (2000): China has 1.25 billion people, has opened to trade, created special economic zones, and accumulated capital. Growth accelerates to 8–10% per capita.

Scale-effect models struggle to explain this: population grew only 25%, so growth shouldn’t have more than doubled. Institutional-improvement models account for it readily: opening markets, improving rule of law, and creating incentives for investment and research drive the acceleration regardless of scale.

See also