Consider an economy with some $n$ agents with continuous utility functions $u_1,\ldots,u_n$. It is easy to prove that a Pareto-optimal allocation exists: define the welfare of an allocation $x$ as: $W(x) := \sum_{i=1}^n u_i(x)$. This $W$ is continuous and the set of allocations is compact, so there exists an allocation maximizing $W$. This allocation must be Pareto-optimal, since every Pareto-improvement would lead to a strictly larger welfare.

When the number of agents in the economy is infinite but countable, a similar proof can be used. Assume that the agents' utility functions are normalized such that all utilities are between $0$ and $1$. Define now $W(x) := \sum_{i=1}^{\infty} 2^{-i} \cdot u_i(x)$. It is possible to prove that $W$ is continuous, so by the same reasoning as above, there is an allocation maximizing $W$, and it is Pareto-optimal.

Now, suppose there is a continuum of agents. One could define a welfare function by e.g. $W(x) := \int_{i=0}^{\infty} 2^{-i} \cdot u_i(x) di$, however, an allocation with maximum $W$ is not necessarily Pareto-optimal: there may exist a Pareto-improvement that gives a higher utility to a single agent, but this will not increase the welfare as this single agent has infinitesimal weight.

Is it possible to prove the existence of a Pareto-optimal allocation in this economy?


1 Answer 1


If (and that is a big if) you have a nonempty compact space of allocations and all utility functions are continuous, then one can prove in a highly nonconstructive way that a Pareto optimum exists.

Theorem: Let $X$ be a nonempty compact space of allocation, $I$ be any nonempty set of agents, and for each $i\in I$ let $u_i:X\to\mathbb{R}$ be a continuous utility function. Then there exists some $x^*\in X$ such that there is no $x\in X$ with $u_i(x)\geq u_i(x^*)$ for all $i\in I$ and $u_i(x)>u_i(x^*)$ for some $i\in I$.

Proof: Choose some well-ordering $<$ of $I$ and construct recursively a transfinite sequence $(X_i)_{i\in I}$ such that $X_i$ is the set of maximizers of $u_i$ in $\bigcap_{j<i}X_j$. Clearly, each $X_i$ is nonempty and compact, provided $\bigcap_{j<i}X_j$ is. The intersection of compact sets is always compact, and it is here nonempty by the finite intersection principle. Every element of $\bigcap_i X_i$ will do as $x^*$. By transfinite induction, every agent has the same utility under every allocation in $\bigcap_i X_i$. To see that these are Pareto optimal allocations, take such an allocation $x^*$ and let $x$ be any allocation under which some agent receives a different utility. Let $i$ be the first such agent. But then $x\in\bigcap_{j<i}X_i$ by transfinite induction. Since $x^*$ maximizes $i$'s utility in this set, we must have $u_i(x)<u_i(x^*)$. $~\Box$

Now, in the classical case of an exchange economy, aggregate feasibility is defined by an integral. The integral of a feasible allocation must equal the integral of the endowment allocation. Since one can always add to what a single agent receives without changing the integral, here, one usually takes the notion of Pareto optimality to be that in every feasible allocation in which a positive measure set of agents is better off, a positive measure set of agents must be worse off (or something similar).

  • $\begingroup$ Thanks! This works, but has the disadvantage that it is not symmetric with respect to the agents, due to the ordering (it is essentially a "serial dictatorship"). Do you know if there is a symmetric solution (such as maximizing the sum in the finite case)? $\endgroup$ Feb 26, 2023 at 14:51
  • $\begingroup$ I'm not sure how to define symmetric allocations in general, but I will think about it. $\endgroup$ Feb 26, 2023 at 18:20

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.