# Competitive equilibrium in Leontief economies

Consider an economy in which all consumers have, possibly different, Leontief utilities. Since preferences are not strictly convex, it is not guaranteed that a competitive equilibrium exists. I found some papers that discuss the computational problem of deciding whether a Leontief economy has a competitive equilibrium, but I am interested in general existence results:

A. What conditions on Leontief economies guarantee that a competitive equilibrium exists?

B. In particular, if the initial endowments are equal (each of $m$ agents receives a fraction $1/m$ of each good), is a competitive equilibrium guaranteed to exist?

• @denesp why did you delete your answer? It almost convinced me... – Erel Segal-Halevi Dec 17 '15 at 18:09
• @denesp Ah, I see! It is an interesting non-example :) – Erel Segal-Halevi Dec 17 '15 at 19:52
• You can try papers on the existence of Nash equilibrium in aggregative games or large anonymous games. A Walrasian economy is such a game (the price vector is the aggregate action) and a Walrasian equilibrium is a Nash equilibrium. Generally existence theorems require compact action sets and continuous utilities. – Sander Heinsalu Jul 4 '17 at 8:07
• It would seem that no true equilibrium exists. only an approximate one when $x_1$ and $x_2$ are continuous. @denesp how does equilibria exist when $p_x=0$? – EconJohn Oct 30 '17 at 17:35
• @EconJohn An example: Let $$U_A(x_1,x_2) = \min(x_1;x_2) \mbox{ and } U_B(x_1,x_2) = \min(x_1;x_2).$$ Assume initial endowments of $(3,2)$ for each player. For any $p_2 \in \mathbb{R}_{++}$ the pricevector $(0,p_2)$ is an equilibrium price vector. This means that given such a price vector each consumer has such an optimal consumption bundle that demand for each good does not surpass supply of respective good. The amount demanded of $x_2$ is trivially $2$ for both players. For $x_1$ it can be any number that is at least $2$. So e.g. $(2,2),(4,2)$ would constitute an equilibrium. – Giskard Oct 30 '17 at 19:10

Arrow-Debreu actually do not just require convexity, they make, as pointed out by denesp in a comment, the convexity assumption (III.c) on utility functions that $u(x)>u(x')$ and $0<t<1$ implies $u(tx+(1-t)x')>u(x')$. Plain convexity suffices for existence, but Leontief preferences do also satisfy condition (III.c).: Assume $\min\{\alpha_i x_i\}>\min\{\alpha_i x_i'\}$. Then $$\min\big\{\alpha_i (tx_i+(1-t)x_i')\big\}>\min\big\{\alpha_i tx_i\big\}+\min\big\{\alpha_i(1-t) x_i'\big\}$$ $$=t\min\{\alpha_i x_i\}+(1-t)\min\{\alpha_i x_i'\}>\min\{\alpha_i x_i'\}.$$