Testing Price Equations

How inflation and unemployment are related in both the short run and long run is perhaps the key question in macroeconomics. This paper tests various price equations using quarterly U.S. data from 1952 to the present. Issues treated are the following. 1) Estimating price and wage equations in which wages affect prices and vice versa versus estimating “reduced form” price equations with no wage explanatory variables. 2) Estimating price equations in (log) level terms, first difference (i.e., inflation) terms, and second difference (i.e., change in inflation) terms. 3) The treatment of expectations. 4) The choice and functional form of the demand variable. 5) The choice of the cost-shock variable. The results suggest that the best specification is a price equation in level terms imbedded in a price-wage model, where the wage equation is also in level terms. The best cost-shock variable is the import price deflator, and the best demand variable is the unemployment rate. There is some evidence of a nonlinear effect of the unemployment rate on the price level at low values of the unemployment rate. Many of the results in this paper are contrary to common views in the literature, but the empirical support for them is strong. ∗Cowles Foundation and International Center for Finance, Yale University, New Haven, CT 06520-8281. Phone: 203-432-3715; email: [email protected]; website: http://fairmodel.econ.yale.edu. For presentation at “The Phillips Curve and the Natural Rate of Unemployment,” June 3–4, 2007, Kiel Institute for World Economics, Kiel, Germany. I am indebted to Jeff Fuhrer and Jesper Lindé for sending me their data.