Introduction: The Question That Dwarfs All Others
Robert Lucas wrote that once one begins to think about long-run growth, it is hard to think about anything else. The reason is arithmetic. A country growing at 1% per year doubles its income in seventy years. A country growing at 3% doubles in twenty-three. Over a century, the difference is not incremental — it is the difference between Malawi and Malaysia.
Robert Solow’s 1956 paper in the Quarterly Journal of Economics, “A Contribution to the Theory of Economic Growth,” provided the framework within which every subsequent growth economist has worked, including those who reject its conclusions. It earned him the 1987 Nobel Prize.
Its central result is unsettling: capital accumulation cannot generate sustained growth in living standards. The engine of long-run prosperity is something the model does not explain.
1. The Building Blocks
The production function
Output is produced by capital and effective labour:
Y = F(K, AL)
where A is labour-augmenting technology (also called labour efficiency). Solow assumes:
- Constant returns to scale. Double both inputs, double output.
- Positive but diminishing marginal products. ∂F/∂K > 0, ∂²F/∂K² < 0. Each additional machine adds less than the last.
- Inada conditions. The marginal product of capital approaches infinity as K → 0 and zero as K → ∞. These guarantee an interior steady state.
The Cobb-Douglas form Y = Kα(AL)1−α satisfies all of these, with α being capital’s share of income — empirically around one-third.
Per-effective-worker terms
Constant returns let us divide through by AL. Define k = K/(AL) and y = Y/(AL). Then:
y = f(k) = kα
This is the single most useful transformation in growth theory. It collapses a two-variable problem into one.
Capital accumulation
A constant fraction s of output is saved and invested. Capital depreciates at rate δ. Labour grows at rate n. Technology grows at rate g.
The fundamental differential equation of the Solow model is:
k̇ = s·f(k) − (n + g + δ)·k
Read it carefully. The first term is actual investment per effective worker. The second is break-even investment — the amount required merely to hold k constant.
Break-even investment has three components:
- δk replaces worn-out capital.
- nk equips new workers with the existing capital ratio.
- gk equips existing workers whose efficiency has risen, so that capital per effective worker holds.
2. The Steady State
Set k̇ = 0:
s·f(k*) = (n + g + δ)·k*
Because f(k) is concave and break-even investment is linear in k, the two curves cross exactly once at a positive k*. The steady state is unique and globally stable.
If k < k*, actual investment exceeds break-even, capital deepens, k rises. If k > k*, the reverse. The economy converges to k* from any starting point.
With Cobb-Douglas, solving explicitly:
k* = [s / (n + g + δ)]1/(1−α)
What grows in the steady state?
This is where students lose marks, so state it precisely:
| Variable | Growth rate in steady state |
|---|---|
| k = K/AL (capital per effective worker) | 0 |
| y = Y/AL (output per effective worker) | 0 |
| Y/L (output per worker — living standards) | g |
| K/L (capital per worker) | g |
| Y (total output) | n + g |
| K (total capital) | n + g |
The steady-state growth rate of output per worker equals g, the rate of technological progress — and nothing else.
3. The Central Result: Saving Does Not Cause Growth
Raise the savings rate s. The s·f(k) curve shifts up. The economy moves to a higher steady state k*. Output per worker rises.
But during the transition, growth is temporarily elevated. Once the new steady state is reached, growth returns to exactly g.
A higher savings rate produces a level effect, not a growth effect.
Nothing a country does to its savings rate, its investment share, or its capital stock can permanently raise its growth rate. Diminishing returns to capital guarantee this. Each additional unit of capital yields less, until it yields exactly enough to cover depreciation and population growth — at which point accumulation stops paying.
This result is the most important thing the model says, and it is why growth economics after Solow became almost entirely about A rather than K.
The Golden Rule
Steady-state consumption per effective worker is c* = f(k*) − (n + g + δ)k*. Maximising over k*:
f′(kgold) = n + g + δ
The marginal product of capital should equal the break-even investment rate. Under Cobb-Douglas this implies a Golden Rule savings rate sgold = α.
An economy with k* > kgold is dynamically inefficient: it could consume more today and forever by saving less. Abel, Mankiw, Summers and Zeckhauser (1989) tested this empirically and found that developed economies sit below the Golden Rule — they save too little, not too much, in this specific sense.
4. Growth Accounting and the Solow Residual
Differentiate Y = Kα(AL)1−α logarithmically:
gY = α·gK + (1 − α)·gL + (1 − α)·gA
We observe gY, gK, gL, and can estimate α from capital’s income share. Everything else is a residual:
Solow residual = gY − α·gK − (1 − α)·gL
This residual is called total factor productivity (TFP) growth. Solow’s own 1957 Review of Economics and Statistics calculation found that the overwhelming majority of US output growth per worker over the first half of the twentieth century was attributable not to capital deepening but to the residual.
Abramovitz memorably called the residual a measure of our ignorance. It contains technology, but also institutional quality, allocative efficiency, human capital not captured in L, measurement error, and everything else the model omits.
5. Convergence: The Model’s Sharpest Prediction
Diminishing returns imply that a country far below its steady state has a high marginal product of capital and therefore grows fast. Poor countries should grow faster than rich ones and catch up.
Absolute convergence
All countries converge to the same steady state. This is empirically false. The cross-country correlation between initial income and subsequent growth is, if anything, positive. The world has diverged, not converged, over the past two centuries.
Conditional convergence
Countries converge to their own steady state, determined by their own s, n, g and δ. Controlling for these determinants, poorer countries do grow faster.
Mankiw, Romer and Weil (1992), in the Quarterly Journal of Economics, tested this directly and produced the most cited paper in empirical growth economics.
Their finding was mixed and instructive. The textbook Solow model, with α = 1/3, could not fit the cross-country data — the observed differences in income were far too large to be explained by observed differences in savings and population growth. Fitting the data required an implausibly high capital share of around 0.6.
Their resolution was the augmented Solow model, adding human capital H as a third accumulable factor:
Y = KαHβ(AL)1−α−β
With physical capital’s share around 1/3 and human capital’s share also around 1/3, the augmented model fit the international data well and delivered a rate of conditional convergence of roughly 2% per year — the famous “iron law of convergence,” implying that a country closes half the gap to its steady state in about thirty-five years.
Crucially, adding human capital raises the total share of accumulable factors from 1/3 to 2/3, which substantially increases how much of cross-country income variation the model can explain, and slows convergence to the observed rate.
The Lucas paradox
If poor countries have low capital and diminishing returns hold, the marginal product of capital there should be enormous. Capital should flood from rich to poor countries.
It does not. Lucas (1990) posed this puzzle, and the answer — differences in A, in institutions, in human capital, in the risk of expropriation — is another way of saying that the Solow model’s exogenous A is doing all the work.
6. What the Model Cannot Explain
Solow’s model explains growth in the steady state by g, and then takes g as given. Technology falls from the sky at a constant rate, identical for all countries, unaffected by any decision anyone makes.
This is not a defect Solow concealed; he stated it explicitly. But it means the model is silent on precisely the thing it identifies as important.
Paul Romer’s response, “Endogenous Technological Change” (Journal of Political Economy, 1990), for which he shared the 2018 Nobel with Nordhaus, made A the outcome of profit-motivated R&D. The key insight is that ideas are non-rival: my using a blueprint does not prevent you from using it. Non-rivalry means ideas have increasing returns at the aggregate level, which breaks the competitive framework — the marginal-cost pricing of a non-rival good yields zero revenue, so innovation requires some form of monopoly rent, typically a patent.
Romer’s model therefore delivers a growth rate that depends on the fraction of resources devoted to research, the size of the population (more people, more ideas), and the institutional structure of intellectual property. Policy can affect long-run growth after all.
The rival research programme, associated with Acemoglu, Johnson and Robinson (2001, American Economic Review), argues that neither capital nor technology is fundamental. Institutions are. Using settler mortality rates as an instrument for the type of colonial institution established, they found large effects of institutional quality on modern income. Their argument is that A differs across countries because property rights, constraints on the executive, and the rule of law differ — and these were shaped by historical accident centuries ago.
7. Exam Technique
AP Macroeconomics
- Draw the diagram: f(k) concave, s·f(k) below it, and (n + g + δ)k as a straight line through the origin. Steady state at the intersection.
- Show that an increase in s shifts s·f(k) up and raises k*, but the economy’s long-run growth rate is unchanged.
- Know that steady-state growth in output per worker equals g.
Cambridge A-Level
- Use the Solow model to evaluate development strategies. A country pursuing growth through investment alone will experience temporary acceleration followed by a return to trend.
- The classic application: the East Asian growth debate. Alwyn Young (1995) and Paul Krugman (1994) argued that Singapore’s extraordinary growth was almost entirely accumulation — rising participation, rising education, rising investment — with negligible TFP growth. Krugman drew the Solow implication directly: such growth must eventually slow, because accumulation runs into diminishing returns. Whether this prediction was borne out remains debated, and it is a superb evaluation point.
- Evaluate: the Solow model treats technology as exogenous, ignores institutions, ignores inequality, and assumes a closed economy with a fixed savings rate. Each assumption is a line of critique.
Summary
The Solow model tells us that capital accumulation raises the level of income but not its growth rate, that long-run growth in living standards is entirely determined by technological progress, and that technological progress is exogenous.
Two of these three propositions are the model’s great achievement. The third is its confession. Sixty years of subsequent research — endogenous growth theory, the institutions literature, the human capital literature — has been an attempt to explain the term Solow could not.
Exercises for Further Thought
1. Mankiw, Romer and Weil rescued the Solow model’s fit to international data by adding human capital, raising the share of accumulable factors from one-third to two-thirds. This is a remarkable improvement — but notice what it accomplishes. By making more of the economy “accumulable,” it attributes more of cross-country income differences to things countries can accumulate, and less to the residual A. Is this an explanation, or a redefinition? Consider: if we kept adding accumulable factors (social capital, organisational capital, institutional capital), we could eventually explain everything and the residual would vanish. Would we then have understood growth?
Suggested reading: Mankiw, N. G., Romer, D., & Weil, D. N. (1992). “A Contribution to the Empirics of Economic Growth.” Quarterly Journal of Economics, 107(2), 407–437. Read alongside Klenow, P., & Rodríguez-Clare, A. (1997), “The Neoclassical Revival in Growth Economics: Has It Gone Too Far?”, NBER Macroeconomics Annual.
2. Krugman and Young argued in the mid-1990s that East Asian growth was “perspiration, not inspiration” — accumulation rather than TFP growth — and therefore must decelerate, as the Solow model requires. South Korea and Singapore have since reached and exceeded the income levels of many Western economies. Was the Krugman-Young thesis wrong, or was it right about the mechanism and wrong about the timing? What evidence over the past twenty years would you examine to decide, and is the distinction between “accumulation” and “technology” even well-defined when accumulating capital is how technology gets embodied?
Suggested reading: Young, A. (1995). “The Tyranny of Numbers: Confronting the Statistical Realities of the East Asian Growth Experience.” Quarterly Journal of Economics, 110(3), 641–680.
References
- Abel, A. B., Mankiw, N. G., Summers, L. H., & Zeckhauser, R. J. (1989). Assessing Dynamic Efficiency: Theory and Evidence. Review of Economic Studies, 56(1), 1–19.
- Acemoglu, D., Johnson, S., & Robinson, J. A. (2001). The Colonial Origins of Comparative Development: An Empirical Investigation. American Economic Review, 91(5), 1369–1401.
- Lucas, R. E. (1990). Why Doesn’t Capital Flow from Rich to Poor Countries? American Economic Review, 80(2), 92–96.
- Mankiw, N. G., Romer, D., & Weil, D. N. (1992). A Contribution to the Empirics of Economic Growth. Quarterly Journal of Economics, 107(2), 407–437.
- Romer, P. M. (1990). Endogenous Technological Change. Journal of Political Economy, 98(5), S71–S102.
- Solow, R. M. (1956). A Contribution to the Theory of Economic Growth. Quarterly Journal of Economics, 70(1), 65–94.
- Solow, R. M. (1957). Technical Change and the Aggregate Production Function. Review of Economics and Statistics, 39(3), 312–320.
- Young, A. (1995). The Tyranny of Numbers: Confronting the Statistical Realities of the East Asian Growth Experience. Quarterly Journal of Economics, 110(3), 641–680.