Two-Sector Growth Models

The economy does not produce a single, undifferentiated good. Two-sector growth models split production into an output-making sector (the “goods” or “final” sector) that produces consumption and capital, and a knowledge-making sector (typically R&D or “research”) that generates new ideas and designs. How the economy allocates labor and capital between these two sectors determines both current consumption and future growth—and this allocation is not automatic.

The Two-Sector Insight: Why Division of Labor Matters for Growth

In a one-sector model, all workers and machines produce consumption goods and capital. The economy reaches a steady state where growth slows or stops unless external technology arrives. In a two-sector model, workers can instead specialize in creating ideas—new designs, processes, medicines, semiconductors architectures—which then raise the productivity of the goods sector.

This creates a deliberate choice: an economy can shift labor toward research, knowingly sacrificing current goods output to raise future output growth. A country that dedicates 3% of its workforce to R&D trades lower current consumption for steeper productivity growth later. One that dedicates only 0.5% accepts slower growth to keep more resources on current production.

This trade-off is not trivial. A young, poor country with urgent consumption needs might rationally choose low research intensity. A wealthy country with satisfied material wants might shift many workers toward innovation. The two-sector model formalizes this logic and shows how it drives long-run divergence between fast-growth and slow-growth economies.

The Romer Framework: Knowledge as an Input

Paul Romer’s 1990 model—the canonical two-sector formulation—separates the economy into three components:

The goods sector. Uses capital $K$ and “effective labor” $A \cdot L_y$ (labor $L_y$ combined with productivity $A$) to make output:

$$Y = K^\alpha (A \cdot L_y)^{1-\alpha}$$

This is a standard Cobb-Douglas production function, but the productivity level $A$ is endogenous.

The research sector. Uses $L_A$ researchers to produce new designs and ideas:

$$\dot{A} = \delta L_A$$

where $\delta$ is research productivity. The more researchers employed, the faster productivity grows. This is the growth engine: a larger research force generates more discoveries, which then compound.

The constraint. Total labor is fixed: $L_A + L_y = L$. Every worker in research is a worker not producing goods.

The economy chooses how to split its labor. High $L_A$ means fast growth but low current output. High $L_y$ means lots of goods now but slower future growth. The steady-state growth rate $g$ is determined entirely by the research labor share: $g = \delta L_A / A$. If the research sector shrinks, growth shrinks. If it expands, growth accelerates.

Steady-State Allocation and Growth

In a two-sector model at steady state, the economy grows at a constant rate $g$. To maintain this rate indefinitely, the labor share in research must stay constant—otherwise growth would accelerate or decelerate. If 20% of workers are always in research and research productivity is stable, growth is stable.

But here is the key: the choice of research intensity is not pinned down by technology alone. Different policies and preferences lead to different allocations. A country with strong patent protection and subsidized universities might sustain 2.5% of its workforce in research. One with weak IP enforcement and low education spending might sustain only 1%. The one with higher research intensity will grow faster, indefinitely.

This is radically different from exogenous growth models, where the growth rate is fixed by outside factors. Here, institutions and policy shape growth. A reform that doubles research efficiency or halves the cost of becoming a researcher raises the long-run growth rate permanently.

Extensions: Multiple Research Sectors

Some two-sector models further subdivide research. For instance, one sector might produce designs for new goods (product innovation) while another improves production methods (process innovation). Or one sector might focus on fundamental science while another commercializes discoveries. The allocation between these research types then becomes a second policy choice.

Other variants nest two sectors vertically: a consumption-goods sector and a capital-goods sector, where the capital-goods sector includes research and development. The split between making consumption goods and making capital (which embodies new technology) drives growth. An economy that invests heavily in capital formation also builds more R&D capacity, creating a reinforcing cycle.

Empirical Patterns: R&D Intensity and Growth

Real-world evidence broadly supports two-sector logic:

• Cross-country correlation. Nations spending 2–3% of GDP on R&D (U.S., Germany, Japan, South Korea) have sustained growth rates of 2–3% per year. Those spending under 1% typically grow at 1–1.5%. The correlation is not perfect—execution matters, as does luck—but the pattern is consistent.

• Sectoral shifts. The U.S. in the 1960s-70s ramped up public R&D spending (space, defense), and productivity growth accelerated. Post-2000, R&D spending moderated in some sectors while booming in IT; productivity growth became uneven by sector.

• Convergence and divergence. Club convergence often reflects research intensity: countries that sustain high R&D intensity converge toward a high-growth club. Those that do not, converge toward a low-growth steady state.

However, not all growth is explained by R&D labor share. Quality of research matters—a researcher in a well-funded lab with modern equipment is more productive than one in an underfunded setting. Spillovers and how freely ideas diffuse also matter. A country can invest heavily in R&D but lose the benefits if knowledge leaks freely to competitors, or gain benefits even with modest spending if it absorbs ideas from abroad.

The Allocation Decision: Market vs. Planner

A key question in two-sector models is whether markets allocate labor correctly between goods and research. In a competitive market, workers flow to whichever sector pays highest wages. If research is less lucrative than manufacturing, workers abandon research, growth slows, and future wages fall. The market might undersupply research because individual researchers do not capture all the benefits of their discoveries—ideas spill over and increase economy-wide productivity, benefiting everyone, not just the discoverer.

This creates a case for subsidizing research beyond the market level: knowledge spillovers mean the private return to research is lower than the social return. Public funding of universities, tax credits for R&D, and patent protection all aim to fix this wedge and sustain more research than markets alone would generate.

Conversely, if research is oversupported by policy—perhaps because of political lobbying—the economy sacrifices too much current consumption and growth does not justify the cost. Finding the right level is an empirical and political question, not a theoretical one.

Growth Implications: Path Dependence and Lock-In

Two-sector models reveal a path-dependence property: the allocation chosen today feeds back into future growth, which shapes future allocations. A country that invests in R&D today builds human capital and knowledge stock, making future research more productive and more attractive. High research today → fast growth → educated workforce → more attractive research → faster growth. This is a virtuous cycle.

Conversely, a low-research country faces a vicious cycle: low research → slow growth → fewer educated workers → research is less productive → little incentive to expand research.

These dynamics suggest that initial conditions and policy choices matter enormously. Two countries with identical technology and preferences but different initial research intensity may diverge persistently. One locks into a high-growth, high-research steady state; the other into a low-growth steady state. This can explain why some regions industrialize rapidly while others stagnate.

Limits and Complications

Two-sector models are highly stylized. Real economies have many sectors, not two, and the boundaries between goods production and research are blurry. A pharmaceutical company both produces drugs and researches new ones; a semiconductor firm improves existing designs while inventing new architectures. Parsing these is empirically messy.

Also, the models often assume research produces a homogeneous “productivity increase,” but in reality discoveries vary wildly in impact. One researcher might improve widget manufacturing by 0.1%; another might invent a transformative technology. Modeling this heterogeneity complicates the theory significantly.

Additionally, human capital—the education and skill of the workforce—plays a large role in both sectors but is often sidelined in basic two-sector models. A low-education workforce cannot sustain effective research, so education policy is inseparable from growth policy.

See also