Choice of Aggregate Demand Proxy and its Affect on Phillips Curve Nonlinearity: U.S. Evidence

Using nonlinearity testing and modeling associated with the smooth transition regression model we examine how the choice of aggregate demand proxy affects, if at all, the nonlinearity of the Phillips curve. The three proxies we examine are the unemployment rate, output gap, and real unit labor costs. Our data is monthly from 1983-2008 for the U.S. We find that regardless of aggregate demand proxy examined, tests indicate a nonlinear Phillips curve. However, the dynamics of the nonlinear models vary substantially.