Mankiw's Puzzle on Consumer Durables: A Misspecification

Mankiw (1982) shows that consumer durables expenditures should follow a linear ARMA(1,1) process, but the data analyzed supports an AR(1) process instead; thus, a puzzle. In this paper, we employ a more general utility function than Mankiw's quadratic one. Further, the disturbance and depreciation rate are respecified, respectively, as multiplicative and stochastic. The analytical consequence is a nonlinear ARMA(infinity,1) process, which implies that the linear ARMA(1,1) is a misspecification. A historical data analysis appears to support the nonlinear model. Since actual data are influenced by historical events, we also carry out a Monte Carlo study to strengthen our point