Paper 4, Section II, H

Optimization | Part IB, 2009

In a pure exchange economy, there are JJ agents, and dd goods. Agent jj initially holds an endowment xjRdx_{j} \in \mathbb{R}^{d} of the dd different goods, j=1,,Jj=1, \ldots, J. Agent jj has preferences given by a concave utility function Uj:RdRU_{j}: \mathbb{R}^{d} \rightarrow \mathbb{R} which is strictly increasing in each of its arguments, and is twice continuously differentiable. Thus agent jj prefers yRdy \in \mathbb{R}^{d} to xRdx \in \mathbb{R}^{d} if and only if Uj(y)Uj(x)U_{j}(y) \geqslant U_{j}(x).

The agents meet and engage in mutually beneficial trades. Thus if agent ii holding ziz_{i} meets agent jj holding zjz_{j}, then the amounts ziz_{i}^{\prime} held by agent ii and zjz_{j}^{\prime} held by agent jj after trading must satisfy Ui(zi)Ui(zi),Uj(zj)Uj(zj)U_{i}\left(z_{i}^{\prime}\right) \geqslant U_{i}\left(z_{i}\right), U_{j}\left(z_{j}^{\prime}\right) \geqslant U_{j}\left(z_{j}\right), and zi+zj=zi+zjz_{i}^{\prime}+z_{j}^{\prime}=z_{i}+z_{j}. Meeting and trading continues until, finally, agent jj holds yjRdy_{j} \in \mathbb{R}^{d}, where

jxj=jyj\sum_{j} x_{j}=\sum_{j} y_{j}

and there are no further mutually beneficial trades available to any pair of agents. Prove that there must exist a vector vRdv \in \mathbb{R}^{d} and positive scalars λ1,,λJ\lambda_{1}, \ldots, \lambda_{J} such that

Uj(yj)=λjv\nabla U_{j}\left(y_{j}\right)=\lambda_{j} v

for all jj. Show that for some positive a1,,aJa_{1}, \ldots, a_{J} the final allocations yjy_{j} are what would be achieved by a social planner, whose objective is to obtain

maxjajUj(yj) subject to jyj=jxj\max \sum_{j} a_{j} U_{j}\left(y_{j}\right) \quad \text { subject to } \sum_{j} y_{j}=\sum_{j} x_{j}

Typos? Please submit corrections to this page on GitHub.