The Solow model has a steady state, and an economy that sits in it is a still picture. This lesson sets it moving. Change the saving rate, the population growth rate or depreciation, or destroy some capital, and watch the economy travel to a new resting point: which things change where it ends up, which change how it grows on the way, and how long the trip takes. Then technical progress is added, which is what finally gives the model growth per worker.
In Lesson 5 the steady state was k* = (s/(n + δ))1/(1−α). That one formula says what every parameter does. Read down the rows below by asking how k* moves, and what that does to output per worker, y* = (k*)α.
| A permanent rise in… | Capital and output per worker, k* and y* | Long-run growth of output per worker | Long-run growth of total output | On the way |
|---|---|---|---|---|
| the saving rate, s | Rise | 0, unchanged | n, unchanged | Growth is faster for a time |
| population growth, n | Fall | 0, unchanged | new n, higher | Growth per worker is slower for a time |
| the depreciation rate, δ | Fall | 0, unchanged | n, unchanged | Growth per worker is slower for a time |
| the level of productivity, A | y* rises in proportion to A1/(1−α) | 0, unchanged | n, unchanged | Growth is faster for a time |
Every row of the “long-run growth” columns is the same, and that is the point. These are level effects: the economy moves to a different resting point, and its output per worker grows faster or slower along the way, but in the long run it is again not growing. A growth effect would change the long-run rate itself. In the basic model nothing inside the model has one.
Two exam statements follow directly. A higher saving rate “raises the growth rate of output in the short run” (the transition) but not the long-run rate (UGC NET 2025). And a higher population growth rate lowers the steady-state capital per worker while raising the steady-state growth rate of aggregate output (JAM 2021, GATE 2026).
This ratio matters for one frequently asked question. UGC NET 2025 lists the effects of faster population growth in the neoclassical model, and the key accepts a fall in capital per head and a rise in the growth rate of total output. The statement that the capital–output ratio decreases is also true by the formula above, since n is in the denominator, yet it is not in the key, so that question is left out of the practice sets below. If you meet a question with this combination, trust the algebra.
Meridia is an invented economy that has sat at its steady state for decades: saving 25%, population growth 1%, depreciation 5%, technical progress 2% and α = 0.33. Choose a shock. The top chart shows the natural logarithm of output per worker, so a straight line is steady growth; the dashed line is where the old parameters would have taken Meridia and the solid line is where it goes now. The lower chart shows the growth rate.
Growth jumps and then fades back to 2%; output per worker ends up higher for good. A level effect, not a growth effect.
Notice the shape of each answer. After the rise in saving, the solid line leaves the dashed line and then runs parallel to it: a level effect. After the rise in technology growth the two lines have different slopes for ever: a growth effect, and the only one. After the earthquake the solid line starts below the dashed line, grows steeply, and rejoins it. The earthquake is a useful test of understanding: because it changes no parameter, it leaves the destination alone and only displaces the economy, so the economy grows faster than normal until the lost capital is rebuilt.
Without technical progress the model has no growth per worker, which contradicts the facts of Lesson 1. Add technology that raises the effectiveness of each worker, so that production is and A grows at a constant rate g. The product AL is effective labour. This is Harrod-neutral technical progress, the only kind compatible with a steady state when growth is constant (Lesson 2).
Now measure everything per unit of effective labour: and . The growth rate of effective labour is n + g, so exactly as in Lesson 5 the growth rate of the ratio is the growth rate of K minus n + g:
At the steady state k̂ and ŷ are constant. But the variables people care about are Y/L = Aŷ and K/L = Ak̂, and these grow because A grows. The whole table of growth rates follows:
| Variable | Steady-state growth rate | Why |
|---|---|---|
| Capital per effective worker, k̂ | 0 | The definition of the steady state |
| Output per worker, Y/L | g | Y/L = A·ŷ with ŷ constant |
| Capital per worker, K/L | g | K/L = A·k̂ with k̂ constant |
| Total output, Y, and total capital, K | n + g | Y = ŷ·AL, with AL growing at n + g |
| Capital–output ratio, K/Y | 0 | K/Y = k̂/ŷ |
| Real wage, w | g | w = A·(ŷ − k̂ f′(k̂)), with the bracket constant |
| Rental rate and profit rate, r | 0 | r = f′(k̂), constant |
Compare this with the list of Kaldor’s stylised facts in Lesson 1: output per worker grows at a steady rate, the capital–output ratio is constant, the profit rate is constant, wages grow with productivity. The Solow model with Harrod-neutral progress reproduces every one of them, which is why it became the benchmark. The price is that the long-run growth rate g is not explained: it is assumed. Explaining it is the work of Lessons 10 and 11.
The steady state says where the economy is going; transition dynamics say how long it takes. Take the growth rate of k̂ and ask how fast it responds to the gap:
This is zero at k̂* and falls as k̂ rises. Differentiate the right-hand side with respect to ln k̂ at the steady state, using s k̂*α−1 = n + g + δ: the slope is −(1 − α)(n + g + δ). So near the steady state the log gap shrinks at a constant rate:
Three things to notice. First, λ does not depend on the saving rate: s decides where the economy is going, not how fast it gets there. Second, a larger α slows convergence, because capital then loses its productivity more slowly, so poor economies get a smaller boost from being poor. Third, the numbers: with α = 1/3, n = 1%, g = 2% and δ = 5%, λ = (2/3)(0.08) = 5.3% a year and half the gap closes in about 13 years. That is fast. Real economies converge much more slowly, about 2% a year in cross-country data. Lesson 8 returns to this gap between the model and the evidence. The exact solution of the equation, which proves the formula for λ, is in the Solow equation, derived and solved.
If convergence took a few years the distinction between levels and growth would hardly matter, because a country that raised its saving rate would reach its new steady state before the next election. The half-life of a decade or more is what makes the transition, with its extra growth, a real part of the story of a country that raises its saving rate.
The passage from one steady state to another has its own name in growth theory: the traverse (Hicks’s term). Steady-state analysis compares the two end points; the traverse is the path between them, and it is analytically awkward because along it neither saving nor the capital–labour ratio sits at its long-run value. UOH 2018 asks for its name.
In discrete time the capital stock evolves by Kt+1 = (1 − δ)Kt + sYt. With labour growing at n this gives kt+1 = [(1 − δ)kt + s f(kt)]/(1 + n), and setting kt+1 = kt = k* gives s f(k*) = (n + δ)k*: the same steady state as in continuous time. The special case δ = 1, in which capital is fully used up within a period, makes Kt+1 = sYt, which is easy to iterate by hand.
Suppose workers save a share sw of wage income and capitalists a share sr of profit income. With Cobb–Douglas production and competitive factor payments, profits are a share α of output and wages 1 − α. Total saving is therefore (sw(1 − α) + srα)Y, an ordinary Solow economy with the average saving rate
For α = 1/3, sw = 0.1 and sr = 0.4 this gives s = 0.1 × 2/3 + 0.4 × 1/3 = 0.2. Raising the capital share α now raises the saving rate when sr > sw, a link between distribution and growth that Kaldor and the Cambridge school made central (Lesson 12).
Questions on what changes in the Solow model, transitions, and the path between steady states. The Delhi School’s full-depreciation and class-saving economies are in Lesson 14 when it opens; the golden rule is the next lesson.
Growth accounting, steady-state algebra, statement questions and matching, up to JNU, IIT JAM and UGC NET level.
This topic’s hardest questions, from the DSE entrance, are worked through in Lesson 14: The steady state and what moves it.
These are questions from the TestStop bank. A free account answers 10 a day, each with its worked solution; All-Access opens every set with no daily limit.
The steady-state capital per worker, output per worker and consumption per worker (up to the golden rule) all rise, so a higher saving rate has a level effect. It does not change the long-run growth rate of output per worker, which is set by technical progress. During the transition to the new steady state, growth is temporarily faster, which is why a question can say both that saving raises growth in the short run and that it does not in the long run.
Faster population growth raises break-even investment, (n + δ)k, because more capital is needed to equip each new worker. Capital per worker and output per worker in the steady state therefore fall. Total output, which grows at the rate of the labour force, now grows faster. So population growth lowers the level of income per head but raises the growth rate of aggregate output.
With labour-augmenting (Harrod-neutral) technical progress at rate g, output per worker grows at g in the steady state and total output grows at n + g. Capital per effective worker, k̂ = K/(AL), is the variable that settles down, and it satisfies k̂̇ = s f(k̂) − (n + g + δ)k̂.
Near the steady state, the gap between capital per effective worker and its steady-state value closes at the rate λ = (1 − α)(n + g + δ). For α = 1/3 and n + g + δ = 8% a year, λ is about 5.3% a year and half the gap closes in about 13 years. The saving rate does not appear in λ, so it changes where the economy is going but not how fast it gets there.
A level effect moves the economy to a higher or lower path for output per worker but leaves the long-run growth rate unchanged: the path shifts in parallel. A growth effect changes the long-run growth rate itself, so the path gets steeper or flatter for ever. In the Solow model, saving, population growth and depreciation have level effects. Only the rate of technical progress has a growth effect, and it is taken from outside the model.
If technology A grows at rate g, then output per worker and capital per worker both keep growing, so there is no constant to find. Dividing by effective labour, AL, gives k̂ = K/(AL) and ŷ = Y/(AL), which do settle down. The steady state is found for these, and the growth rates of the original variables are recovered by adding g back: per worker, g; in total, n + g.