The Solow Growth Model

Why growth needs its own model

The classical model treats capital K and labor L as given: Y = F(K, L). It does not track how today’s saving becomes tomorrow’s capital. The Solow model adds that missing piece, a law of motion for capital, and uses it to answer three questions:

  • Why do economies with little capital grow fast, and why does that growth slow down (postwar Japan)?
  • Does saving more make a country grow faster forever, or only richer?
  • What is the only source of sustained growth in living standards?

The building blocks

Everything is measured per worker (lower-case letters), using constant returns to scale:

Piece Equation Meaning
Production y = f(k) Output per worker rises with capital per worker k = K/L, at a diminishing rate (MPK falls as k rises)
Saving = investment i = s·y = s·f(k) A fixed share s of output is saved and invested; c = (1 − s)·y is consumed
Depreciation δ·k Capital that wears out each period
Law of motion Δk = s·f(k) − δ·k Capital per worker grows when investment exceeds what is needed to replace worn-out capital
Steady state s·f(k*) = δ·k* Investment exactly covers break-even investment, so Δk = 0

With population growth n and labor-augmenting technology growth g, measure everything per effective worker (k̃ = K/(E·L)) and break-even investment becomes (δ + n + g)·k̃:

  • δ·k̃ replaces worn-out capital,
  • n·k̃ equips new workers,
  • g·k̃ keeps up with rising labor efficiency E.

These are the small-rate approximations used in the lecture (exact in continuous time).

1. The Solow diagram

Try it. The defaults are the lecture’s numerical example: y = √k (α = 0.5), s = 0.30, δ = 0.10, n = g = 0, so k* = 9, y* = 3, c* = 2.1. Move one slider at a time and watch where the investment curve s·f(k) crosses the break-even line.

f(k): output
s·f(k): investment
(δ+n+g)·k: break-even investment

What to notice

  • Saving (s up): the blue curve shifts up and k* moves right. Output per worker y* is permanently higher. That is a level effect.
  • Depreciation or population growth (δ or n up): the red line gets steeper and k* moves left. Faster population growth means lower output per worker, because each period’s investment is spread over more new workers (capital dilution).
  • Golden Rule: turn it on. Steady-state consumption c* = f(k*) − (δ+n+g)k* is largest where the slope of f(k) equals the slope of the break-even line: MPK = δ + n + g. With Cobb–Douglas output this happens exactly when s = α. In the lecture example, k*GR = 25, y*GR = 5, c*GR = 2.5 (above c* = 2.1 at s = 0.30).
  • The two extremes: at s = 1 everything is saved, so c = 0. At s = 0 nothing is invested, capital wears away, and since a worker with no capital produces nothing (f(0) = 0), output and consumption go to zero in the long run. Steady-state consumption is zero at both ends, so the best saving rate lies in between: the Golden Rule. (Try s = 0 in the table in section 3 below.)
  • Growth rates: whatever you do to s or n, long-run growth of Y/L is g. Only technology moves the long-run growth rate.

Why we need the Golden Rule

The Solow diagram shows that saving more always raises steady-state capital and output. That raises a natural question: should a country just save as much as possible? No. People care about what they consume, not about how much capital or output exists. More saving builds capital but leaves a smaller share of output to consume (c = (1 − s) × y). We need a rule that picks the saving rate that is best for consumption in the long run. That rule is the Golden Rule: the saving rate that gives the most consumption per worker in the steady state.

Why such a rule must exist. Long-run consumption is zero at both extremes:

  • s = 1: everything is saved, so nothing is consumed.
  • s = 0: there is no investment, capital wears away, and output goes to zero (see the worked example in section 3).

Consumption is positive in between, so some saving rate in the middle gives the most. The Golden Rule finds it.

The condition. Steady-state consumption is c* = f(k*) − δk*. It is largest where one more unit of capital adds exactly as much output as it costs to maintain:

  • MPK = δ in the basic model.
  • MPK = δ + n + g once you add population growth and technology growth.

On the diagram, that is the point where the slope of f(k) equals the slope of the break-even line. Shortcut: with Cobb–Douglas output y = kα, the Golden Rule saving rate is just s = α.

Why it matters for policy. Comparing MPK with δ + n + g tells a country which way to move. Below the Golden Rule (MPK > δ + n + g) it should save more, which costs consumption today but raises it later. Above the Golden Rule (MPK < δ + n + g) it should save less, which raises consumption at every date. A high saving rate on its own does not tell you which case a country is in. Section 5 runs this check for the U.S.

2. The transition: what happens after a policy change

The diagram shows where the economy ends up. This simulation shows the path it takes. The economy starts in the steady state at the old saving rate; in period 10 the saving rate changes permanently.

Fixed: α = 0.5, δ = 0.10, n = 0. Uses the lecture's approximation k̃t+1 = k̃t + s·f(k̃t) − (δ+n+g)·k̃t. Grey dashed path = no policy change.

What to notice

  • Level effect, not growth effect. Raise s from 0.30 to 0.40 and look at Growth rate of Y/L: growth jumps above g, then falls back to g. The economy ends up permanently richer but does not grow permanently faster.
  • Below the Golden Rule (old s < 0.50): switch to Consumption per worker. Consumption drops at period 10, then climbs above the old path. The current generation sacrifices for future ones.
  • Above the Golden Rule (try old s = 0.70, new s = 0.50): consumption jumps up immediately and stays above the old path at every date. Everyone gains.
  • Convergence. Growth is fastest right after the change, when k̃ is furthest below its new steady state. This is the model’s reading of postwar Japan.

3. The lecture’s numerical example, period by period

The defaults are the lecture’s example: y = √k, s = 0.30, δ = 0.10, n = g = 0 and k₁ = 4, which gives k* = 9. Change the inputs and the table updates. The last row is the steady state for whatever inputs you enter.

At the defaults, start k₁ above 9 (say 16) and Δk is negative: the economy shrinks back toward the same k* = 9. In general, any k₁ above k* shrinks toward it and any k₁ below grows toward it.

What if nobody saves? (s = 0)

Set s = 0 and k₁ = 9 in the table above. The law of motion becomes Δk = 0 · f(k) − δk = −δk: no new investment, and 10% of capital wears out every year.

Year k y = √k c = y (all consumed)
1 9.00 3.00 3.00
2 8.10 2.85 2.85
3 7.29 2.70 2.70
10 3.49 1.87 1.87
30 0.42 0.65 0.65
long run → 0 → 0 → 0
  • Consumption looks great at first: 3.00 in year 1, versus 2.1 when s = 0.30, because nothing is set aside.
  • But that consumption comes from running down the capital stock. Capital falls 10% a year (9 → 8.1 → 7.29 …) toward zero.
  • Output per worker depends on capital per worker, and a worker with no machines produces nothing: f(0) = 0. So output and consumption also go to zero.
  • Steady-state consumption is c* = f(0) − δ·0 = 0, the same as at s = 1. That is why the Golden Rule saving rate lies strictly between 0 and 1.

4. Growth accounting: the Solow residual

For Y = A·Kα·L1−α, output growth splits into capital, labor, and total factor productivity (TFP):

ΔY/Y ≈ α·ΔK/K + (1 − α)·ΔL/L + ΔA/A

TFP growth ΔA/A is what is left over: the Solow residual.

The residual picks up everything not measured in K and L: genuine innovation, but also changes in capacity utilization, organization, reallocation across sectors, and measurement error.

5. Is the U.S. below the Golden Rule?

The lecture’s stylized calculation uses three ratios to output. Compare MPK − δ with n + g: if MPK − δ is larger, the economy has too little capital (below the Golden Rule).

Balanced growth: who grows at what rate

Variable Symbol Steady-state growth rate
Capital per effective worker k̃ = K/(EL) 0
Output per effective worker ỹ = Y/(EL) 0
Capital per worker K/L g
Output per worker Y/L g
Total capital K n + g
Total output Y n + g

Convergence: conditional, not unconditional

  • Unconditional convergence (all poor countries grow faster than rich ones) is not what Solow predicts, and the cross-country data do not show it.
  • Conditional convergence: a country grows faster the further it starts below its own steady state. Countries with different saving, population growth, or technology have different steady states.
  • Low income alone does not imply fast growth. What matters is income relative to the country’s own long-run path.

Check yourself

1. The saving rate permanently doubles from 20% to 40%. Does output per worker grow faster forever? No. Growth of Y/L rises during the transition to the new, higher k*, then returns to g. The permanent effect is on the level of output per worker, not its growth rate.
2. Population growth n rises. What happens to k*, y*, and the growth rate of total output Y? The break-even line (δ+n+g)k gets steeper, so k* and y* per worker fall (capital dilution). Total output Y grows faster in steady state, at n + g.
3. y = √k, δ = 0.10, n = g = 0. Find the Golden Rule capital stock and saving rate. MPK = 0.5/√k = δ = 0.10, so √k = 5 and k*GR = 25. Then y = 5, break-even investment = 0.10 × 25 = 2.5, so s_GR = 2.5/5 = 0.5 (equal to α), and c*GR = 5 − 2.5 = 2.5.
4. An economy is above the Golden Rule. Should it save more or less, and who gains? Less. Consumption jumps up immediately and stays higher along the whole transition, so every generation gains.
4b. A country saves nothing (s = 0). Why does its long-run consumption fall to zero even though it consumes all of its output? With no investment, Δk = −δk: capital only depreciates and is never replaced, so k falls toward 0. Since f(0) = 0, output falls toward 0, and consumption (which equals output) falls with it. Steady-state consumption is f(0) − δ·0 = 0.
5. Output grows 3%, capital 4%, labor 1%, α = 0.3. What is TFP growth? 3 − 0.3(4) − 0.7(1) = 3 − 1.2 − 0.7 = 1.1%.
6. Two countries both have income per worker of $5,000. One grows at 5% a year, the other at 0.5%. Does this contradict the Solow model? No. Each converges to its own steady state. The fast grower is likely far below its own k* (high saving, good institutions, technology adoption); the slow grower may already be near a low k*.
7. Which raises long-run living standards more: a permanent doubling of s, or a permanent 1-point rise in g? The rise in g. Higher s gives a one-time level increase; higher g raises the growth rate of Y/L forever, so it eventually dominates any one-time level gain.

What the Solow model leaves unexplained

The model takes s, n, and g as given. It does not explain where technological progress comes from, why saving rates differ across countries, or how human capital and institutions (compare North and South Korea) shape investment and innovation. It is also a real model: it says nothing about money, the price level, or inflation.