Let's take the example of a generic Harrod-neutral (labor-augmenting) production function $f(k)$; all letters denote the growth rates they usually would. In the regular Solow growth model with the Inada assumptions, it is a result of its comparative statics that, since $sf(k^*)=(n+\delta+g)k^*$ is the steady-state condition, if we assume $k^*$ is a function of the savings rate then $\frac{\partial k^*}{\partial s}>0$. Total differentiation gives us that $$ \frac{\partial k^*}{\partial s} = \frac{f(k^*)}{(n+g+\delta)-sf'(k^*)},$$ and in my class the professor claimed that this last expression is strictly positive because the numerator is (fine) and the denominator is too. Basically, it boils down to the fact that $\frac{n+g+\delta}{s} > f'(k^*)$, and that this is a fact implied by the Inada assumptions.

I have tried messing around with the statement about the limiting behavior of the derivative at 0 (infinity) and as k tends to infinity (zero); all to no avail. If someone could please offer some hints or an explanation or a proof of this fact I would deeply appreciate. This was not a question posed for an assignment or anything of the sort, just personal interest (having some background in analysis).

  • $\begingroup$ Do you make the assumption that $f$ is concave? $\endgroup$
    – Giskard
    Sep 30, 2017 at 20:41
  • $\begingroup$ Yes, it is one of the Inada conditions. $\endgroup$
    – spacetime
    Oct 1, 2017 at 22:48

1 Answer 1


We want to prove that

$$\frac{n+g+\delta}{s} > f'(k^*)$$

Replace the left hand side with the equivalent from the expression $sf(k^*)=(n+\delta+g)k^*$, and you get:

$$ \frac{f(k^*)}{k^*} > f'(k^*) $$

Cobb-Douglas case

Without loss of generality, assume $$f(k^*) = {k^*}^{\alpha}$$

Then, the above inequality is:

$$ {k^*}^{\alpha-1} > \alpha{k^*}^{\alpha-1} $$

This is:

$$ 1> \alpha $$

This condition represents a production function with decreasing marginal returns to capital, necessary for the Inada conditions to hold.

General case

A concave function has the following property, for any $x$ and $y$ in the domain:

$$f(y) \leq f(x) + f'(x)(y-x) $$

Rearranging, this is:

$$ \frac{f(y)- f(x)}{y-x} \leq f'(x) \hspace{1cm}, \text{for } y>x $$

This is, the derivative of $f(x)$ at point $x$ is no smaller than the slope of the segment between $x$ and $y$. For the case of a strictly concave function (which is needed for the Inada conditions to hold), the inequality is strict ($<$), except in the trivial case of $x = y$.

For further confirmation, apply the above when $y=0$ and $x=k^*$ (notice here $y<x$ so the sign of the above inequality reverses). Since $f(0)=0$, you get:

$$ \frac{f(k^*)}{k^*} > f'(k^*) $$

which is what we wanted to demonstrate.

  • 1
    $\begingroup$ Very nice. Everything makes sense; I just have one lingering concern. Why can you assume WLOG that the function takes on some Cobb-Douglas form? I had heard from someone that given the Inada conditions the production function basically has to be "asymptotically Cobb-Douglas", perhaps it has to do with that? Would love some clarification! Many thanks $\endgroup$
    – spacetime
    Oct 1, 2017 at 22:51
  • $\begingroup$ @spacetime Well, you mentioned the "regular Solow growth model", which is based on a Cobb-Douglas. Is that not the case? A general CES also holds the Inada Conditions. The proof is more complex then. $\endgroup$
    – luchonacho
    Oct 2, 2017 at 7:16
  • $\begingroup$ @luchonacho That is definitely not a WLOG situation. You have to make that assumption in the beginning. You can explain that it is customary, but it does come with loss of generality. $\endgroup$
    – Giskard
    Oct 2, 2017 at 11:09
  • $\begingroup$ @denesp it is WLOG within Cobb-Douglas. Personally, I have never seen a Solow model with a PF other than CD; hence my interpretation of "regular Solow growth model". If the OP would have been clearer, we could get what s/he meant in the first place. $\endgroup$
    – luchonacho
    Oct 2, 2017 at 11:16
  • $\begingroup$ It is true, I was thinking of the "regular" Solow growth model as one with a CD production function. However my broader interest, especially after @luchonacho's answer, was in taking any arbitrary function with constant returns to scale that satisfies the Inada conditions (these en.wikipedia.org/wiki/Inada_conditions). Apologies for the confusion and thank you for the thorough answer. $\endgroup$
    – spacetime
    Oct 2, 2017 at 16:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.