Suppose that $F(\cdot)$ has CRS in $K$ and $L$, the elasticity of Substitution is $\sigma_{K L} \equiv F_{L} F_{K} / F F_{L K}$.

I once derived this equation but I remember that it takes me quite long time and the process is quite tedious. I wonder is there any easy way to derive this or remember it intuitively?


2 Answers 2


I am not sure if it is intuitive but this is because because CRS function is homogenous of degree 1.

Full derivation:

First, general formula for any arbitrary elasticity of substitution between $L$ and $K$* is given by (see Sydsæter et al. EMEA pp 430):

$$\sigma_{L,K} = \frac{-F_K'F_L'(xF_K'+ yF_L')}{xy \left( (F_L')^2F_{KK}^{''} - 2 F_K'F_L'F_{KL}'' + (F_K')^2F_{LL}''\right)}, \text{ for: } F(K,L)=c \tag{1}$$

Where c is arbitrary constant. This is our starting point.

Now by definition property of CRS production function is that they are homogenous of degree 1 (since by definition we have CRS when $F(tK,tL)= tF(K,L)$).

If $F$ is homogenous of degree 1, then the numerator of (1) will be $= −F_K'F_L'F$.

This is because of the Euler's theorem which tells us that if:

$$f(x,y) \text{ is homogenous of degree k} \implies xf_x'(x,y) + yf_y'(x,y)= kf(x,y)$$

Next Euler theorem also implies that (assuming $f$ is twice continuously differentiable) that:

$$xf_{xx}''(x,y) + yf_{yx}''(x,y)= (k-1)f_x'(x,y)$$ $$xf_{yx}''(x,y) + yf_{yy}''(x,y)= (k-1)f_y'(x,y)$$

The above implies that in our case:

$$K F_{KK}'' = - LF_{KL}'' $$ and that $$L F_{LL}'' = - K F_{LK}'' = - K F_{KL}'' $$. Hence the denominator will be given by:

$$ -F{KL}''\left( L^2 (F_L')^2 + 2KLF_K'F_L'+ K^2(F_K')^2\right)= - F_{KL}''(KF_K'+ L F_L')^2 = - F_{KL}''F^2$$

We use Euler theorem above again.

Now finally we are done:

$$\sigma_{LK} = \frac{−F_K'F_L'F}{- F_{KL}''F^2} =\frac{F_K'F_L'}{ F_{KL}''F} $$

I personally do not think the result above is very intuitive (if it is so its intuition eludes me), but it is a consequence of CRS function being homogenous of degree 1 and this result actually applies for any elasticity of substitution between two variables of arbitrary function that is homogenous of degree 1 and twice continuously differentiable.

* or $y$ and $x$ for that matter this generalizes to any elasticity of substitution problem

  • 1
    $\begingroup$ Yes, what confuse me is that this result should be so general but neither its derivation nor the result itself quite intuitive. Anyway, thanks for your answer and I think it's nice to have an answer that contains the full derivation in this site. $\endgroup$ Commented May 16, 2021 at 21:17
  • $\begingroup$ @Alalalalaki yea I don't really think there is an easier derivation, not all results in economics need to be intuitive, I mean this is why we model stuff, sometimes we learn something non-intuitive yet tremendously useful. $\endgroup$
    – 1muflon1
    Commented May 16, 2021 at 21:19

1muflon1's answer is entirely correct.

Let me give another alternative derivation, which might be a little bit easier to remember, although probably not more intuitive.

Let $k = K/L$ be the capital to labour ratio. Then we can define $f(k) = F(K/L, 1)$ to be the output per unit of labour. then by the CRS assumption we have: $$ F(K,L) = L f(K/L) = L f(k). $$ Taking derivatives with respect to $K$ and $L$ gives: $$ \begin{align*} &F_K = L f'(k) \frac{1}{L} = f'(k),\\ &F_L = f(k) + L f'(k) \left(-\frac{K}{L^2}\right) = f(k) - k f'(k). \end{align*} $$ Finally: $$ \begin{align*} &F_{K,L} = f''(k)\left(-\frac{K}{L^2}\right) = -kf''(k) \frac{1}{L},\\ \to &F_{K,L} F = -k f''(k) f(k). \end{align*} $$ Let's now have a look at the IES: $$ \sigma_{L,K} = -\frac{\partial \ln(F_K/F_L)}{\partial \ln (K/L)} = \frac{\partial \ln \left(\frac{f'(k)}{f(k) - k f'(k)}\right) }{\partial \ln k} $$ Now take derivative of both numerator and denominator with respect to $k$: $$ \begin{align*} \sigma_{L,K} &= -k \frac{f - k f'}{f'}\frac{f'' \left(f - k f'\right) - f'\left(f' - f'- k f''\right)}{(f - k f)^2}\\ &= \frac{-k f'' f}{f'(f - kf)},\\ &= \frac{F_{K,L} F}{F_K F_L} \end{align*} $$


Your Answer

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

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