This paper attempts to survey and appraise various contributions to the microtheoretic foundations of Keynesian economics. Samuelson's neoclassical synthesis asserts that the neoclassical price theory is valid when full employment is achieved by the Keynesian device for economic stabilization. Lately this contention has been challenged by such counter-counter-revolutionists as Clower, Leijonhufvud, and others, including even a neoclassical theorist like K. J. Arrow. We shall examine the pioneering works on this line of reappraisal, and make clear why it has farreaching consequences on the whole structure of conventional equilibrium economics.
First we shall discuss the consequences of removing one of the main assumptions of the Walrasian theory, that of the "tâtonnement" process. When transactions are allowed at nonequilibrium prices, the usual concepts of notional supply and demand are no longer valid, and they have to be replaced by those of effective supply and demand. This implies that traders will modify their supplies and demands according to the perceived constraints, as pointed out by Clower, Don Patinkin, R. J. Barro, and H. I. Grossman. Many relations in Keynesian economics are given a theoretical foundation within the framework of this "dual decision hypothesis, " and this idea has further been generalized by Jean-Pascal Benassy so that it can be incorporated into the true general disequilibrium model with multiple traders and markets.
We shall next discuss the consequences of removing another basic assumption of the Walrasian theory, that of the "market auctioneer." Once the fictitious auctioneer is banished out of our picture, somebody will have to take over the task of price determination so as to personalize the impersonal law of supply and demand in some way. This leads us to examine Franklin Fisher's work on price adjustment without an auctioneer, in which "dealers" adjust their own prices, and "customers" search among the dealers to find that dealer whose price is most advantageous for them. Another work of much interest in this context is a joint paper by E. S. Phelps and S. G. Winter, Jr., which incorporates the noninstantaneous "customer flow" dynamics. This implies that customers shift gradually through the process by which information is transmitted over time.
These models assume naturally that price can be diversified in a market of even a physically well-defined commodity. However, it should be noted that this is a case of some differentiated products, for, as Iwai pointed out, if the commodity is homogeneous and all the shops are located at one spot, there can be hardly any problem of searching costs.
Finally we shall append the discussion on price adjustments versus quantity adjustments, as described by Leijonhufvud. Though the extreme Keynesian assumption of infinitely fast quantity adjustments yields a recursive model close to the typical multiplier process, it is not clear to what extent its "full employment counterpart" is different from the traditional equilibrium theory with the usual price response to excess demands.

View full abstract

Frequently economic models are given in the form of discrete linear system, that is, in the form of simultaneous linear difference equations. In this paper, we shall study some characteristics of this system, and show that some concepts of the control theory can be introduced into the field of econometrics. At the same time, we shall see how a certain foundamental problem of the control theory is to be solved through simple methods which are familiar to all econometricians.
The mathematical description of the discrete linear system is given through following state equations; X(t+1)=AX(t)+CU(t) Y(t)=BX(t)+DU(t) Vector X(t) stands for the set of state variables at time t, vector U(t) for the level of inputs at t, and vector Y(t) for the level of outputs at t.
In this paper, state equations are generalized as X(t+1)=AX(t)+CU(t)+Z(t) Y(t)=BX(t)+DU(t)+ W(t) Random vector Z(t) and W(t) constitute the levels of disturbances at time t. If we allow this generalization, new problems must be solved. Among them are observability and controllability conditions of the discrete linear system. In this paper, we shall show that some of these problem can be solved through econometric approach.

View full abstract