Welcome to your study space
Understand economics at your own pace: step-by-step guided courses, and exercises to practice.
The Solow model: saving and the steady state
Why a country that saves twice as much does not grow twice as much — and why saving 70% makes you poorer.
Δk = s·f(k) − (n + δ)·kProblem / motivation
If saving finances machines, and machines finance output, then a country that saves more should grow richer faster — indefinitely. Why do governments not push the saving rate to 50%, to 70%, to 100%?
The question is not absurd: it has been asked for real. China raised its investment to nearly 50% of its GDP in the 2000s — a level Europe never came close to — and yet its growth has slowed sharply since. The model built by Robert Solow in 1956, which won him the Nobel prize in 1987, explains exactly why, and its answer fits in one sentence: accumulating capital runs into two walls, one that slows you down, one that costs.
This course will not merely state the result. It builds the equation that governs accumulation, PROVES that the economy inevitably comes to rest at one precise and unique point — the steady state — then establishes a second, still more disturbing result: beyond a certain saving rate, a country produces more while consuming LESS. So there is an optimal saving rate, and it is not 100%.
Assumptions
In your view, which assumptions are needed for this model to hold? Jot down your ideas — no lead is wrong, this is your worksheet.
Your worksheet is still empty. Go for it: propose at least one idea.
0 idea(s) proposedFormalization
CAPITAL here is not money: it is machines, buildings, roads, computers — everything that has been produced and that serves to produce. Write K for that stock for the whole country, L for the number of workers, and **k = K/L** for capital per worker. Thanks to constant returns to scale (assumption 1), output per worker depends on it alone. The algebra of that step takes one line: if doubling both factors doubles output, then dividing both by L divides output by L — that is, F(K, L)/L = F(K/L, 1). The right-hand side no longer depends on anything but k, and we write it f(k). Hence **y = f(k)**. We shall take f(k) = √k, a square root: it rises, but more and more slowly — going from 1 to 4 machines takes output from 1 to 2, but going from 4 to 9 takes it only from 2 to 3. There are the diminishing returns, in two numbers. (The general form is written k^α; α measures the share of capital in value added, about 1/3 in the data. We take α = 1/2 so that every computation comes out round, and the simulator will let you test 1/3.)
What becomes of the capital stock from one year to the next? Two forces pull against each other. What INCREASES it: investment. Households save a share s of their income, and that saving becomes investment (assumption 3): this adds s·f(k) per worker. What REDUCES it: first wear and tear — a share δ of the stock goes to rust and obsolescence each year, i.e. δ·k. Then demography: if the population grows at rate n, the same stock has to be shared among more heads, which lowers capital PER WORKER by about n·k — the capital has not disappeared, it has been diluted. The two add up into a single maintenance requirement, (n + δ)·k. Hence the fundamental equation of the model, the one that was missing: **Δk = s·f(k) − (n + δ)·k**. In words: capital per worker rises by what is invested, minus what it takes to maintain what exists. The whole course consists in reading that line.
Compare the two terms. The first, s·f(k) = s·√k, rises while flattening: that is the concavity of assumption 2. The second, (n + δ)·k, is a straight line through the origin: the maintenance requirement is proportional to the stock. A curve that flattens and a line that rises steadily necessarily cross once — and once only for k > 0. (They also meet at k = 0, but that point does not interest us: it is the economy without a single machine, which produces nothing and can invest nothing. Set it aside and only one crossing is left.) That point, written k*, is the STEADY STATE: there, Δk = 0 and capital stops moving. It remains to show that we get there, and the sign of Δk says so. Take s = 20% and δ = 5% (with n = 0 for simplicity). At k = 4: you invest 0.20 × √4 = 0.40, you must maintain 0.05 × 4 = 0.20 — investment wins, Δk = + 0.20, capital RISES. At k = 25: you invest 0.20 × 5 = 1.00, you must maintain 0.05 × 25 = 1.25 — maintenance wins, Δk = − 0.25, capital FALLS. Below k* we go up, above it we come down: the point is unique, and it is STABLE. Whatever its starting point, the economy lands there.
Let us first solve for k*. Setting Δk = 0 gives s·√k* = δ·k*. We can divide by √k* — non-zero, since k* = 0 is the other solution, that of an economy with nothing — and what is left is s = δ·√k*, so **k* = (s/δ)²** and y* = s/δ. With s = 20% and δ = 5%: k* = 4² = 16 and y* = 4. Double the saving, s = 40%: k* = 64, y* = 8. Capital quadruples, output doubles: more saving does give a higher level. But what do you CONSUME? Not y*, only what is left once maintenance is paid: c* = y* − δ·k*. At s = 20%: 4 − 0.8 = 3.2. At s = 50%: k* = 100, y* = 10, maintenance 5, so c* = 5.0. At s = 70%: k* = 196, y* = 14, maintenance 9.8, so c* = 4.2 — less than at 50%, for a clearly higher output. Three points are not a proof, so let us prove it in two lines, with no derivative at all. Substitute: c* = y* − δ·k* = s/δ − δ·(s/δ)² = (s − s²)/δ. Recognise a parabola in s. Complete the square in the numerator: s − s² = − (s² − s) = − (s − ½)² + ¼. So **c* = [¼ − (s − ½)²] / δ**. Since the square is always positive or zero, something is always taken away from ¼ — except at one point: s = ½, where the square vanishes. The maximum is therefore reached exactly at **s = ½ = α**, and it is worth 1/(4δ) = 5. This is the GOLDEN RULE, and you can see at once why it is symmetric: saving 30% or 70%, twenty points either side, gives strictly the same consumption. Beyond it, the country works for its machines. Click each term:
Solving / calculation
Three passes: the march towards the steady state year by year, the effect of doubling saving, then the golden rule. The toy is f(k) = √k, δ = 5%, n = 0 — chosen so the arithmetic comes out round; the real orders of magnitude are given alongside.
- A country starting poor (k₀ = 1, s = 20%)
Δk = 0.20·√1 − 0.05·1 = 0.20 − 0.05Δk = + 0.15: it equips itself fast - The same country, once well equipped (k = 9)
Δk = 0.20·3 − 0.05·9 = 0.60 − 0.45Δk = + 0.15 again, but on a stock 9 times bigger - Further still (k = 15)
Δk = 0.20·3.873 − 0.05·15 = 0.775 − 0.75Δk = + 0.025: the braking is clear - ⭐ The stop: Δk = 0 gives k* = (s/δ)²
k* = (0.20/0.05)² = 4²k* = 16 · y* = √16 = 4 - Checking stability on either side
at k = 4: 0.40 > 0.20 · at k = 25: 1.00 < 1.25we rise below k*, we fall above it - Growth is crushed (path from k₀ = 1)
growth of y: ≈ 5.2%/yr over the 1st decade → 1.7% around year 30 → 0.6% around year 60it tends towards ZERO, not towards a positive plateau - Doubling saving (s = 40%)
k* = (0.40/0.05)² = 64 · y* = 8level doubled — but long-run growth still nil - ⭐ The golden rule: what can be CONSUMED, c* = y* − δ·k*
s = 20% → 3.2 · s = 50% → 5.0 · s = 70% → 4.2maximum at s = α = 50%; beyond that, the country grows poorer
The model therefore answers the opening question twice over. (1) Saving more raises the long-run LEVEL of living, never its GROWTH RATE: going from 20% to 40% doubles output per worker, but once the new plateau is reached growth falls back to zero. The reason is the wall of diminishing returns — each extra machine yields less, while maintenance grows in proportion. (2) There is even a point beyond which saving more makes you POORER: at 70%, this country produces 14 and consumes only 4.2 of it, against 5.0 when saving 50%. That is what feeds the debate on Chinese over-investment — nearly 50% of GDP invested, when the share of capital in value added, which sets the golden rule, is closer to a third. That leaves a puzzle: if accumulation always runs out of steam, where does the growth observed over two centuries come from? The model has an answer, and it is an uncomfortable one (interpretation, point 2).
Δk = s·k^α − (n + δ)·k; steady state k* = (s/(n+δ))^(1/(1−α))This simulator follows an economy YEAR BY YEAR from its starting point, instead of displaying only where it ends up. Set the saving rate and watch two things: the growth of output per worker, which is crushed whatever you do, and long-run consumption, which passes through a maximum. Push saving to 70% to see a country grow poorer while producing more.
Economic interpretation
Three readings: what the model forbids you to hope for, where growth then comes from, and what it predicts for poor countries.
This is the distinction that structures the whole model, and that public debate confuses endlessly. A lasting rise in the saving rate moves the economy onto a higher path: more capital per worker, more output per worker, permanently. But only during the transition is growth faster; once the new steady state is reached, it falls back exactly where it was — to zero per worker in this model without technical progress. So a country cannot settle durably at 5% growth by saving ever more: each rise in saving buys a plateau, never a slope. And the simulator shows that these transitions are long: the speed of convergence is (1 − α)(n + δ). Do the sum with realistic values — α = 1/3, n = 1%, δ = 5%: (1 − 1/3) × 6% = 2/3 × 6% = 4% a year. At that pace it takes about seventeen years to close HALF the gap, and more than half a century for the bulk of it. A whole generation can therefore live in transition without ever seeing the steady state.
If accumulation runs out of steam, how can living standards have been multiplied more than tenfold in two centuries? Solow's answer is honest and disturbing: through TECHNICAL PROGRESS, which has to be added to output in the form y = A·f(k), where A measures how efficiently capital and labour are combined. That A escapes the wall of diminishing returns, because it does not accumulate like a stock of machines: an idea can be used everywhere at once without wearing out. At the steady state, output per worker then grows at the rate of A, and total GDP at the rate of A plus population. The trouble is that the model says nothing about what makes A advance: it puts it in from outside. Solow himself measured it in 1957 — nearly 87% of the growth of output per hour worked in the United States between 1909 and 1949 was attributable to it (course “Growth accounting”). So the main engine of growth is, in this model, precisely what it does not explain: it is that observation which launched, thirty years later, the theories of endogenous growth.
A direct consequence of diminishing returns: where capital is scarce it yields a great deal, so it accumulates fast. Two countries sharing the same parameters — same saving rate, same demography, same institutions — therefore converge towards the SAME steady state, with the poorer one growing faster. That is CONDITIONAL convergence, and the word matters: it is conditional on the parameters. Nothing predicts that all countries meet, only that each reaches ITS own plateau. The data bear it out in part — the spectacular catch-up of Japan, then Korea, then China came with very high investment rates — but also refute it: countries with little capital have stagnated for decades, which the model attributes to different parameters, without explaining why they differ.
Limits / critiques
Exercises
Model of the course (f(k) = √k, n = 0). With s = 25% and δ = 5%, what is capital per worker at the steady state, k*?
Still with δ = 5%, the economy is at k = 36 and the saving rate is 25%. What does capital do?
With s = 50% and δ = 5%, long-run output is y* = 10 and capital k* = 100. What is long-run consumption per worker?
True or false: in this model, a country that saves more always ends up consuming more.
Two countries have the same parameters (s, n, δ, technology), but one has capital per worker twice as low. What does the model predict?
Same situation (s = 25%, δ = 5%). The economy is at k = 9. What is Δk that year?
The saving rate goes from 25% to 50% (δ = 5%, n = 0). By how much does long-run OUTPUT per worker rise (in %) — careful, we ask about output y*, not capital k*?
A very poor country starts from k = 1 (s = 20%, δ = 5%, n = 0). What is its capital per worker after ONE year?
True or false: the Solow model proves that capital accumulation CANNOT, on its own, explain the growth of living standards over two centuries.