growing, there has been an increasing clamor from certain quarters for the reverse. The cry is for small models, not big ones, and for simple models, not complex ones. It is almost as if all that matters is prediction of current dollar and real gross national product (GNP and GNP58) and perhaps a few of their components. But surely this is blindness to the demands of government policy-makers and business forecasters for more and more specific detail. Some policy and business decisions are based on aggregate phenomena. However, in both cases, it is the impacts on individual sectors, industries, groups and products that are of most concern and not the level of GNP. Those who argue for highly aggregated, small models seem to want them for the sake of simplicity. What this entails, of course, is that many of the equations of such models are reduced form or quasi-reduced form predictive devices rather than structural relationships. One is reminded of the conception of modeling the economy held by some members of the Chicago school. Supposedly, since the supply of money is almost all that matters, then, at least according to their view, all that needs to be done is to make real or nominal GNP a function of the money supply. Never mind that the causal mechanism is incompletely spelled out and that it is impossible to test the hypothesis except by observing a few statistical significance coefficients. This reliance on predictive rather than structural relationships is one of the principal disadvantages of small models. Such models also are subject to aggregation errors, particularly in the areas of price, industrial production, employment, investment and inventory functions. On the other hand, small models have the