Under what conditions constraints start to bind and how to find it

I was trying the following optimization problem:

$$ \mathscr{L} = x_1 x_2 + x_2 + \lambda(M-P_1 x_1-P_2 x_2) + \mu x_1$$

The thing is, as you can see, I operate with $U = x_1 x_2 + x_2$, which implies that $x_1$ can be zero for a problem to provide an interesting solution. That is why I have included the "non-negativity" condition for $x_1$ in Lagrangian. I have performed the Lagrangian according to the classic approach and gained the following results:

\begin{align*} &(1) &&& &(2) \\ \mu &= 0 &&& \mu &>0 \\ x_1 &= \frac{M-P_1}{2 P_1} &&& x_1 &= 0 \\ x_2 &= \frac{M-P_1}{2P_2} + \frac{P_1}{P_2} &&& x_2 &=\frac{M}{P_2} \end{align*}

These are the demands for each case of $\mu$... However, what I would like to know is when each case becomes relevant? For which prices and income is relevant $(2)$ instead of $(1)$ etc...

The thing is during computation I came to the following condition (in case of $\mu > 0$):

$$ x_2 + \mu = P_1 \frac{1}{2} $$

However, when solving graphically, there were multiple cases in which $\mu>0$ and they depended on both the prices and the income.

Can I find out for which prices $\mu = 0$ and for which ones $\mu > 0$? How can I systematically find for which prices the constraint binds and for which ones it does not, without relying on graphical solution?


1 Answer 1


For problems involving non-negativity constraints, you’d solve the usual Lagrangian without them and check if the optimal variable(s) ended up negative or not.

If they end up negative, we set them $= 0$.

The Lagrangian would be

$\mathcal{L} = x_1 x_2 + x_2 + \lambda (M - P_1 x_1 - P_2 x_2)$

Solving this yields the demands

$x_1^\star = \frac{M - P_1}{2 P_1}$

$x_2^\star = \frac{M + P_1}{2 P_2}$

Note the expression for $x_1^\star$ can be negative, this happens when $P_1 > M$.

This implies that if $P_1 \leq M$, the marshallian demands are the $x_1^\star, x_2^\star$ we got above.

If $P_1 > M$, then we set $x_1^\star = 0$.

Then we get $x_2^\star = \frac{M}{P_2}$.

By the way, your method yielded the correct demands for both cases.

In your method, what you would do is check the case where the non-negativity constraint is not binding $(\mu = 0)$.

If the non-negativity constraint is not violated (the expression for $x_1^\star$ is not negative), that is your solution.

If it is violated (the expression for $x_1^\star$ is negative), the other case is your solution.


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.