Characterization of intertemporal optimality in terms of decentralizable conditions: The discounted case (Q1110433)

The paper studies the problem of characterizing optimal paths in the context of a Ramsey type multisector capital accumulation model where the welfare function, technology and discount factor are stationary over time. The problem in brief is \(_{t=0}\delta^ tW(c_ t)\) subject to \(0\leq c_ t\leq y_ t\) for all t, \((x_ t,y_{t+1})\in \Omega\) (the technology set) for all t and \(c_ t=x_ t+y_ t\) all t, \(y_ 0=\tilde y\). It is well-known from earlier works that optimal paths can be characterized by the existence of present value prices \(p_ t\) for the goods in each period t such that (a) the given program is ``competitive'' i.e., ``profits'' including utility as a final good is maximized along the program over the feasible initial and final stocks: \(\delta^ tW(c_ t)-p_ tc_ t\geq \delta^ tW(c)-p_ tc\) all \(c\geq 0\), all \(t\geq 0\), \(p_{t+1}y_{t+1}-p_ tx_ t\geq p_{t+1}y-p_ tx\) all (x,y) in \(\Omega\), all \(t\geq 0\) and (b) transversality condition is satisfied i.e., Lim \(p_ tx_ t=0\) where \(x_ t\) stands for the initial stocks in period t, \(y_ t\) output and \(c_ t\) consumption in period t. The paper shows that within the class of competitive paths an alternative characterization is possible if there is a stationary optimal path with capital stock \(x^*\) and stationary prices \(p^*\). Under some weak regularity condition on \(x^*\) it is established that a competitive program is optimal if and only if the following condition is satisfied: (c) \((p_ t-p^*)(x_ t-x^*)\leq 0\) for all t. The difference in the two characterization results is that condition (b) is replaced by (c). This has certain advantages. Condition (b) is an asymptotic condition which can be verified only ``at infinity''. In static optimality problems price characterization results provide ``decentralizable'' rules of decision making which, if followed by myopic agents who have only partial information, results in the optimal solution. A condition like (b) is not of such a type while condition (c) is decentralizable in this sense. Myopic agents can verify such a rule on a period by period basis along with profit maximization which results in an optimal solution. Another interesting implication is that according to condition (c) a suboptimal program can always be detected within some finite horizon. This is in contrast to condition (b) according to which optimality or suboptimality can only be detected by analyzing the asymptotic behaviour of \(p_ tx_ t.\) The regularity condition on \(x^*\) required to obtain the above result is that there is a \(\lambda >1\) such that \((x^*,\lambda x^*)\) is a feasible input output pair.