A Comparision of Eleven Econometric Models of the United States.

This paper compares solution error and dynamic multipliers of major U.S. econometric models. The models have been simulated within sample periods, from fixed initial conditions, for intervals up to eight quarters in duration and for a long continuous span of time. Where possible, the simulations have been confined to the uniform time span 1961 through 1967. The errors are relatively low in comparison to levels and changes in the variables and are indicative of strong performance. Errors in simulation of first differences of gross national product (GNP) and related variables reveal a stable and small error pattern. First-difference errors do not exhibit strong growth trends. For most models this stability is more pronounced for real rather than nominal GNP, indicating difficulties of accurately predicting prices. Given nonlinear relationships between real output and prices, increments in nominal or constant dollar fiscal stimulus will reveal different multiplier responses at various stages of the economy's growth cycle. At low utilization rates real multipliers will be greater than when capacity is more fully utilized. Much of the sustained multiplier increase is attributable to pressures on prices, which seem to accelerate as simulation periods are lengthened. At the end of five to ten years, some of the models show that continued sustained fiscal stimulus has ever-increasing perverse impacts.