Asymmetric time series and temporal aggregation

The detection of nonlinearities could depend on the sampling frequency. Asy mmetric monthly series may become symmetric when aggregated to quarterly or annual frequencies. We test against nonlinearity using the nonlinear autor egressive asymmetric moving average (ARasMA) model, which nests the linear ARMA model as a special case. Using monthly, quarterly, and annual Swedish unemployment series, we find support for symmetry/linearity in the annual s eries but not in the monthly and quarterly series.