Consider the following problem \begin{equation*} \begin{aligned} & \underset{x_1, x_2}{\text{maximize}} & & u(x_1, x_2) \\ & \text{subject to} & & p_1x_1 + p_2x_2 = y \end{aligned} \end{equation*} Suppose I found optimal values $(x_1^*, x_2^*)$ that maximize my function $u$.

Now consider problem

\begin{equation*} \begin{aligned} & \underset{x_1, x_2}{\text{maximize}} & & v(u(x_1, x_2)) \\ & \text{subject to} & & p_1x_1 + p_2x_2 = y \end{aligned} \end{equation*}

Does the optimal value change?

First order conditions (FOCs) for the first problem are \begin{equation*} \begin{aligned} & \frac{\partial u}{\partial x_1} - \lambda p_1 = 0 \\ & \frac{\partial u}{\partial x_2} - \lambda p_2 = 0 \end{aligned} \end{equation*}

For the second one

\begin{equation*} \begin{aligned} & \frac{\partial v}{\partial u}\frac{\partial u}{\partial x_1} - \lambda p_1 = 0 \\ & \frac{\partial v}{\partial u}\frac{\partial u}{\partial x_2} - \lambda p_2 = 0 \end{aligned} \end{equation*}

Not sure how to proceed from here. One way to find an example of $v$ that changes optimal value. I tried several candidates but get same optimal values. A link or description of relevant material will be helpful.

  • $\begingroup$ Set $v(x)=-x$. Then clearly a maximum for $u$ is a minimum for $v$ and conversely. But in general you shouldn't expect any relation between the two.. $\endgroup$ – b00n heT Sep 9 '16 at 9:31
  • $\begingroup$ @b00n heT Thanks for your comment. What if I specify that $v$ is a positive monotonic transformation? Such as for example $v = x^2$ or $ v = e^x$? Then there will be a relation? $\endgroup$ – tosik Sep 9 '16 at 9:34
  • $\begingroup$ The relation must be monotonic increasing. $\endgroup$ – Yves Daoust Sep 9 '16 at 9:35
  • $\begingroup$ Of course there will be: if some function is monotone, then maximizing/minimizing the function is equivalent to maximizing/minimizing its argument. $\endgroup$ – b00n heT Sep 9 '16 at 9:36

If $v$ is strictly growing on the domain where $u$ take its values, it doesn't change the optimums

If $v$ is strictly decreasing on the domain where $u$ take its values, it turns minimas into maximas, and maximas into minimas

Rq : This is often exploited to simplify problems

Following comment : I don't think this is worth calling theorem, and I don't think there is a special name for this property. The proof : (for one case)

Let us consider a set $S$, $f$ a function on $S$ such that the maximum of $f$ on $S$ exists, $\max_{x \in S} f(x) = x_0$, and $g$ a growing fuinction on $f(S)$.

$\forall x \in S, f(x) \leq f(x_0)$ so $\forall x \in S, g(f(x)) \leq g(f(x_0)) $

  • $\begingroup$ Thanks for your answer. It will be really helpful if you could give me a name or a link to a theorem that ensures that positive monotonic transformation doesn't change optimal values $\endgroup$ – tosik Sep 9 '16 at 9:55
  • $\begingroup$ Ahh, so obviously, thanks $\endgroup$ – tosik Sep 9 '16 at 12:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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