Adjustment of Economic Time Series for Outliers, Seasonal, and Calendar Effects

The article tests a procedure for spotting and eliminating statistical outliers in economic time series which does not require previous information on the location of outliers and which estimates model parameters and outliers simultaneously. Results for three different time series suggest that this approach may be an alternative to intervention models. Furthermore, allowing for calendar and outlier effects significantly reduces the variance of model-based seasonal adjustment procedures and may prove clearly superior to conventional empirical-technical methods (Census X-11).