Macroeconometric model comparison and evaluation techniques: a practical appraisal

This paper examines four model comparison techniques in the setting of an actual large-scale U.S. macroeconomic model. The techniques include formal hypothesis tests, mean squared error analysis, and two new tests based on the complementary conditions that the structural and reduced form disturbances should be orthogonal to the predetermined variables. The latter two tests can be conveniently implemented by regressing either structural residuals or reduced form forecasting errors on a vector of predetermined variables. The empirical example tests restrictions on the simultaneous feedback effects of prices, wage rates, and interest rates.