The estimation and use of a quadratic input-output model for macroeconomic analyses of the U.S. economy

Abstract

For the past fifty years, the government has attempted to direct and control the growth of the U.S. economy through a variety of policy tools. These efforts have intensified in recent years. Concurrently, the economy has become increasingly complex. The production processes in most sectors have become more dependent on the output of other sectors. That is, the ability of any given sector to produce a given level of output increasingly depends on the ability of other sectors to provide needed goods and services to be used as input in this sector's production process. Because of these interdependencies, any proposed policy should be studied to determine its overall effect on the economy, and not just that narrow segment of the economy for whose benefit it was designed. Due to the interdependencies between sectors and capacity constraints placed on the output of any given sector, particular policy change may not have the desired effect. This study provides a model useful in such analyses. The objective of this study is to postulate an interdependent model of the U.S. economy containing those explanatory variables relevant for such an analysis, estimate the coefficients of this model, and demonstrate its usefulness by investigating the sector-by-sector effects of an action by agricultural producers which reduced the total output of raw agricultural products. This objective was accomplished by first postulating a model of the U.S. economy which accounts for own-price, cross-price, income and foreign exchange effects on the final demand for goods from eight major production sectors in the U.S. economy. The model was developed in an input-output framework in which production functions are fixed proportion and final demand includes consumer demand, net inventory changes, net capital formation, net export demand, and government demand. Coefficients in the final demand equations for seven of these sectors were then estimated using the two-stage least-squares estimator...