Solow Steady-State Capital per Effective Worker

Robert Solow's 1956 paper in the Quarterly Journal of Economics asked what happens to an economy that saves a constant fraction of its output and invests it in capital that depreciates. The answer is the most important negative result in growth economics, and this formula is where it lives.

The mechanism is a race between two flows. Investment adds capital, and because output has diminishing returns to capital — that is what $\alpha < 1$ does — the additions get smaller as the capital stock grows. Meanwhile, three things dilute capital per effective worker at a constant proportional rate: depreciation $\delta$ wears it out, population growth $n$ spreads it over more workers, and technological progress $g$ raises the number of "effective" workers each real worker counts as. Investment is a curve that flattens; dilution is a straight line. They cross once, and $k^*$ is where.

The result: saving more raises the LEVEL of income per worker, and does not raise its long-run growth rate. An economy that lifts its saving rate grows faster for a while as it climbs to a higher steady state, and then settles back to growing at exactly the rate it grew before. This was a genuinely unwelcome finding in 1956, when raising the investment rate was the standard prescription for development, and it remains the model's central lesson.

The steady state is a limit the economy never reaches. Convergence is asymptotic: an economy below $k^*$ approaches it and arrives after infinite time. Empirical estimates put the convergence speed near 2% of the remaining gap per year, meaning a half-life of roughly thirty-five years. Since $s$, $\delta$, $n$ and productivity all move on shorter timescales than that, the target shifts before the approach is finished. What the model describes is a direction of pull, not a place anyone has been.

And now the honest problem, which is the reason the model is taught rather than used. The only thing that makes income per worker grow in the long run is $g$ — technological progress — and $g$ is EXOGENOUS. It arrives from outside the model, unexplained, at a rate the model simply assumes. Solow's own 1957 growth accounting found that the majority of US output growth per worker could not be attributed to capital deepening at all; it was residual. So the model's central variable is precisely the one it does not account for, and Solow said so plainly himself. The endogenous growth literature of the 1980s and 1990s — Romer, Lucas and the rest — exists to try to put $g$ inside the model, with mixed success.

Two more cautions on the arithmetic here. All three rates must be per the SAME period, because the model adds them, and nothing in a units engine can check that: a depreciation rate per year added to a population growth rate per decade produces a confident and meaningless answer. And $k^*$ is capital per EFFECTIVE worker, not per worker — the two differ by the level of technology, and along the balanced growth path capital per actual worker keeps rising at rate $g$ forever even though $k^*$ is constant.

One result worth knowing that this page does not compute: the golden-rule saving rate, the one that maximizes consumption per worker in the steady state, is exactly $\alpha$. Saving more than that buys a bigger capital stock at the price of consuming less forever. A higher $k^*$ is not automatically better, which is a useful corrective to reading this formula as an instruction.