Optimal intertemporal allocation mechanisms and decentralization of decisions (Q1110436)

When the planning horizon is unbounded and the number of agents infinite intertemporal profit maximization does not guarantee efficiency [\textit{E. Malinvaud}, Econometrica 21, 233-268 (1953; Zbl 0052.160)]. An additional condition such as ``terminal cost minimization'' or a ``transversality condition'' becomes indispensible. The problem with these additional assumptions is - as it has been pointed out by \textit{T. C. Koopmans} (1957) - that one cannot imagine how they could be sustained by ``decentralized'' decision making. The objective of this paper is to conceptualize the idea of informationally decentralized mechanisms in an infinite horizon model in which decision makers are finite-lived. The results are essentially impossibility theorems. One theorems states ``that, provided the class of a priori admissible environments is sufficiently rich, it is impossible to design a temporally decentralized mechanism that is both non-wasteful and unbiased''. (The latter two concepts are based on the notion of efficiency rather than optimality, and do not involve the knowledge of preferences.) Similar impossibility results are obtained if the ``discounted sum of felicities'' is maximized subject to time-varying non-stationary technologies. ``It is impossible to design a temporally decentralized mechanism guaranteeing optimality.'' There is only one noteworthy exception, namely, if the felicity function is logarithmic and the technologies linear.