The ESM Procedure
Missing Value Modeling Issues
The treatment of missing values varies with the forecasting model. Missing values after the start of the series are replaced with one-step-ahead predicted values, and the predicted values are used in the smoothing equations.
The treatment of missing values can also be specified with the SETMISSING= option, which changes the missing values prior to modeling.
Note: Even if all of the observed data are nonmissing, the ACCUMULATE= option can create missing values in the accumulated series (when the data contain no observations for some of the time periods specified by the INTERVAL= option).