I present a decentralized process for finding the equilibria of an economy characterized by a finite number of linear equilibrium conditions. The process finds all equilibria or, If there are none, reports that, in a finite number ofsteps at most equal to the number of equations. The communication and computational complexity compare favorably with other decentralzed processes. The process may also be interpreted as an algorithm for solving a distributed system of linear equations. Comparisons with the Linpack program for LU (lower and upper trianula decomposition of the matrix of the equation system, a version of Gaussian elation) are presented.This paper presents a decentralized process for finding the equilibria of an economy characterized by a finite number of linear equilibrium conditions. The process differs from others that have been studied in connection with stability (global and local) of equilibrium in that it is given by an algorithm requiring a finite number of steps, rather than an infinite iterative process, as, for instance, a process given by difference or differential equations.Furthermore, it is not assumed that equilibria exist. Ifthere are no equilibria, the process discovers this and stops; ifthere are equilibria, the process finds them all in a fixed number of steps depending on the number of equilibrium conditions.The process is decentralized in the following sense. In a certain subset of economies with N agents, the equilibrium condition of an individual agent is a linear equation in RN (possibly several linear equations)t with N + 1 coefficients, not all zero.t That equation (i.e., its coefficients) need be known only to the agent whose equilibrium condition it is. The process requires each agent to transmit a message consisting of a certain finite number of points of RN to one other agent. When N > 2, the messages transmitted are insufficient for any agent to learn the complete system of equilibrium equations. § At each step ofthe process one agent decides on a message by using only knowledge ofhis/her own equilibrium condition and the message received from one other agent.Literature. The problem of finding equilibrium is related to and stems from previous work on the stability of equilibrium (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13). [Work on computation of equilibria should also be mentioned, although that deals with nonlinear equilibrium conditions and does not concern itself with decentralized methods (14).] For the present purpose the literature on stability of equilibrium may be summarized as follows. Papers (1-3) studied price adjustment processes in which prices change according to excess demand. The objective was to show that every economic environment that satisfies conditions ensuring the existence of competitive equilibria has (globally or locally) stable equilibria. This turned out not to be the case (4, 5). To guarantee stability, additional conditions, such as "gross substitutes," are required. Smale (6) studied an adjustment process that involved much bigge...