A method of principal components is employed to investigate nonlinear dynamic factor structure using a large panel data. The evidence suggests the possibility of nonlinearity in the U.S. while it excludes the class of nonlinearity that can generate endogenous fluctuation or chaos