Solow Growth Model

The setup

The model has one good. Output Y is produced from capital K and labor L using a Cobb-Douglas production function with constant returns to scale and a capital share α between zero and one:

Y = Kα L1−α

Convert to per-worker units by dividing through by L. Let k = K/L and y = Y/L. Then:

y = kα

A constant fraction s of output is saved and invested. Capital depreciates at rate δ per year, and the labor force grows at rate n. The capital stock per worker evolves according to:

Δk = s·kα − (δ + n)·k

The first term is investment per worker. The second is what you'd lose anyway — depreciation plus the dilution from a growing labor force, since adding a new worker without new capital lowers the per-worker stock. The economy is in steady state when these two balance: Δk = 0.

Worked example — solving for the steady state

Take the canonical textbook calibration: capital share α = 1/3, savings rate s = 0.20, depreciation δ = 0.05, labor-force growth n = 0.01.

Set Δk = 0:

s·k*α = (δ + n)·k*
0.20 · k*1/3 = 0.06 · k*

Solve for k*:

k*1−α = s / (δ + n)
k*2/3 = 0.20 / 0.06 = 3.333
k* = 3.3333/2 ≈ 6.085

Steady-state output per worker is:

y* = k*α = 6.0851/3 ≈ 1.826

Steady-state consumption per worker (income minus saving) is:

c* = (1 − s) · y* = 0.80 · 1.826 ≈ 1.461

Now ask the policy question: what if the country raised its savings rate to s = 0.30? Repeat:

k*2/3 = 0.30 / 0.06 = 5.000
k* = 53/2 ≈ 11.180
y* ≈ 2.236
c* = 0.70 · 2.236 ≈ 1.565

Output per worker rose by 22%. Consumption per worker rose by only 7%, because much of the extra output goes to replacing the larger stock of capital. The transition from the old to the new steady state is gradual — a half-life of roughly 25 years for typical parameters — which is why "raise the savings rate" works in textbooks but is slow in practice.

Convergence and the catch-up effect

Two countries with the same s, δ, n, and α should converge to the same y*. A country starting below its steady state grows faster than one near it, because the marginal product of capital is larger when k is small. This is the conditional convergence prediction.

The data are mixed. East Asian "tigers" — Japan, South Korea, Taiwan, Singapore — converged toward U.S. levels at exactly the rate Solow predicts. Sub-Saharan Africa did not, but the model says it shouldn't have, because the parameters differ: lower savings, higher depreciation through political instability, lower technology.

The augmented Solow model of Mankiw, Romer, and Weil (1992) added human capital h to the production function:

y = kα hβ

and showed that with α + β ≈ 2/3, conditional convergence fits roughly 80% of the cross-country variance in per-capita income. Their paper became the most-cited empirical defense of the Solow framework.

The golden rule

Saving more raises output but also raises the capital you must replace. There's a savings rate that maximizes steady-state consumption per worker. Set up the maximization:

max c* = f(k*) − (δ + n)·k*

The first-order condition is f′(k*_gold) = δ + n. The marginal product of capital equals the rate at which capital must be replaced. For Cobb-Douglas, f′(k) = α·k^α−1, so:

α · kgoldα−1 = δ + n
kgold1−α = α / (δ + n)

For α = 1/3, δ = 0.05, n = 0.01: k_gold^2/3 = 0.333/0.06 = 5.556, so k_gold ≈ 13.09. The golden-rule savings rate equals α — for α = 1/3, that's 33%, well above empirical U.S. saving of around 20%. The data say most rich countries save below the golden rule, which the model treats as suboptimal.

Solow model vs other growth theories

	Solow (1956)	Harrod-Domar (1939–46)	Endogenous growth (Romer 1990)	Augmented Solow (MRW 1992)	Schumpeterian (Aghion-Howitt 1992)
Source of long-run growth	Exogenous technology	Capital accumulation	R&D, human capital	Exogenous tech + h-capital	Quality-ladder innovation
Production function	Cobb-Douglas Y = KαL1−α	Leontief (fixed proportions)	Y = Kα(AL)1−α, A endogenous	Y = KαHβL1−α−β	Variable productivity
Diminishing returns to K	Yes	No (knife-edge)	Avoided via spillovers	Yes	Yes per vintage
Convergence prediction	Conditional	None — instability	None — divergence possible	Conditional, fits data better	Conditional
Policy levers	Saving rate, n	Saving, capital-output ratio	R&D subsidies, IP, education	Saving + education	Innovation incentives
Empirical fit	Decent within country	Poor	Good for cross-country divergence	Best fit MRW found	Mixed

The Solow residual

Solow's 1957 follow-up paper applied the model to U.S. data, 1909–1949. He decomposed output-per-worker growth into two parts:

Δy/y = α · Δk/k + Δa/a

Δk/k he could measure from capital data. The total Δy/y came from national accounts. Whatever was left — Δa/a — became known as the Solow residual or total factor productivity (TFP). The headline number: of the roughly 1.5% per year per-worker growth, only about 0.2 points came from capital deepening. The remaining 1.3 points were the residual, which Solow interpreted as technological progress.

This was a startling result. It said the economy's long-run improvement comes overwhelmingly from things the model treats as exogenous — better engineering, organization, ideas — and only a small fraction from saving and investment. It set the agenda for half a century of growth research.

Counterarguments

The endogenous growth critique. Romer (1986, 1990) and Lucas (1988) argued that long-run growth shouldn't be a free parameter. Knowledge is producible, and once produced, it doesn't depreciate the way capital does — it accumulates. Endogenous growth models replace the exogenous A with R&D-driven knowledge accumulation and explain why innovative countries grow faster.

The unit-of-time problem. The model treats time as continuous and steady states as attainable. Critics — including Solow himself — note that the half-life of convergence is so long (decades) that the steady state is more conceptual than predictive over policy-relevant horizons.

Aggregation. A single capital stock K combines tractors, factories, software, and skyscrapers. The Cambridge capital controversy of the 1960s argued that aggregating heterogeneous capital using market prices is logically circular when those prices depend on the aggregate. The mainstream sidesteps this by pragmatic assumption.

The development critique. The model says any country can converge to U.S. income by saving enough and waiting. Empirically, most poor countries don't. Acemoglu, Johnson, and Robinson have argued the missing variable is institutions — property rights, rule of law, corruption — which the Solow model treats as fixed background.

Common pitfalls — endogenous vs exogenous

• Calling Solow an "endogenous growth model" because saving is endogenous in some extensions. The defining feature is whether technology is endogenous, not whether saving is. Even with optimal saving (the Ramsey-Cass-Koopmans extension), Solow remains exogenous-growth because A is given.
• Confusing the level effect with the growth effect. Raising the savings rate raises the long-run level of y* but doesn't raise its growth rate — that's still pinned to the rate of technological progress. This is the cleanest way to test whether you understand the model.
• Reading the steady state as a destination. Real economies are constantly being shocked off their steady states. The steady state is a moving attractor, not a place countries reach.
• Treating the Solow residual as "technology." The residual is anything not explained by measured K and L — it includes mismeasurement, scale effects, allocative efficiency, and yes, technology. Calling it pure technology overstates what the data say.
• Forgetting depreciation in the per-worker accounting. The break-even line is (δ + n)·k, not just δ·k. Population growth dilutes capital just as surely as wear and tear, and skipping the n term leads to wrong steady-state numerics.