N a note in Econometrica,2 in I942, the author sketched a new approach to the theory of business fluctuations. He claimed that this was the approach in a mathematical sense, as defined by Jeffreys.3 It also can be considered as the simplest possible dynamic generalization of the Walrasian system of general equilibrium. Roughly speaking, the simple theory explains business fluctuations as resulting from the existence of interrelated markets. The buyers and sellers on these markets react not to existing prices but to anticipated prices. All functional relationships involved are assumed to be linear (as first approximations). If anticipations of the buyers and sellers are asymmetrical in a certain well defined sense, then there is the possibility of periodic fluctuations in prices and quantities. These magnitudes will fluctuate with the same period but with various amplitudes (which may be constant, damped, or exploding) and with leads and lags. Such a movement of prices and quantities is essentially what we mean if we speak of the business cycle.4 It is not claimed, of course, that this is in all cases the only possible explanation of booms and depressions. The author has indicated in an earlier publication that a variety of causes may be at work in creating these fluctuations.5 But it is possible that these speculative fluctuations are able to explain some of the cyclical phenomena met with in our economy.6 Strictly speaking, our theory would require a study of all or at least the most important markets. Such a study is not feasible because of the amount of computational work, which, as will be seen, is very great even with only three variables. We, therefore, shall limit ourselves to the most important markets and shall use for our verification data for stock prices, farm prices, and prices of nonfarm commodities, covering the period I920-42. All the series are in the form of index numbers; we hope that this fact does not unduly distort the picture. A model based upon our theory will be constructed and the unknown parameters determined by various methods, keeping always in mind the relationships necessarily implied by our theory.7 The fitted functions will be subjected to certain statistical tests in the final sections of this paper.8 Such a procedure cannot, of course, afford a complete test of the theory. Even if the statistical test were completely applicable, the model used is not unique for our theory. All that can be said is that ours is the simplest theory which will lead to the specific model.